[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-he-y--Awesome-Pruning":3,"tool-he-y--Awesome-Pruning":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",140436,2,"2026-04-05T23:32:43",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":76,"owner_avatar_url":77,"owner_bio":78,"owner_company":79,"owner_location":80,"owner_email":81,"owner_twitter":81,"owner_website":82,"owner_url":83,"languages":81,"stars":84,"forks":85,"last_commit_at":86,"license":81,"difficulty_score":87,"env_os":88,"env_gpu":89,"env_ram":89,"env_deps":90,"category_tags":93,"github_topics":94,"view_count":23,"oss_zip_url":81,"oss_zip_packed_at":81,"status":16,"created_at":99,"updated_at":100,"faqs":101,"releases":142},2086,"he-y\u002FAwesome-Pruning","Awesome-Pruning","A curated list of neural network pruning resources.","Awesome-Pruning 是一个精心整理的神经网络剪枝资源合集，旨在为深度学习模型压缩领域提供一站式导航。随着深度神经网络日益庞大，如何在保持精度的同时减少计算量和存储占用成为关键挑战，而剪枝技术正是解决这一问题的核心手段。Awesome-Pruning 通过系统性地收录从 2015 年至今的前沿论文、代码实现及技术综述，帮助从业者快速掌握滤波器剪枝、权重剪枝等不同类型的主流方法。\n\n该资源库特别适合人工智能研究人员、算法工程师及高校学生使用。无论是希望了解“彩票假设”等理论基础的研究者，还是寻求高效模型部署方案的开发者，都能在此找到对应的学术成果与开源代码。其独特亮点在于不仅按年份和会议（如 ICLR）分类整理文献，还提供了一个结构化的剪枝技术分类图谱，并收录了作者团队发表在 IEEE TPAMI 上的结构化剪枝权威综述，为用户构建了清晰的技术演进脉络。作为社区驱动的项目，它持续更新最新研究成果，是进入神经网络剪枝领域不可或缺的入门指南与参考手册。","# Awesome Pruning [![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re)\r\n\r\nA curated list of neural network pruning and related resources. Inspired by [awesome-deep-vision](https:\u002F\u002Fgithub.com\u002Fkjw0612\u002Fawesome-deep-vision), [awesome-adversarial-machine-learning](https:\u002F\u002Fgithub.com\u002Fyenchenlin\u002Fawesome-adversarial-machine-learning), [awesome-deep-learning-papers](https:\u002F\u002Fgithub.com\u002Fterryum\u002Fawesome-deep-learning-papers) and [Awesome-NAS](https:\u002F\u002Fgithub.com\u002FD-X-Y\u002FAwesome-NAS).\r\n\r\nPlease feel free to [pull requests](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fawesome-Pruning\u002Fpulls) or [open an issue](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fawesome-Pruning\u002Fissues) to add papers.\r\n\r\n## Table of Contents\r\n\r\n- [Type of Pruning](#type-of-pruning)\r\n\r\n- [A Survey of Structured Pruning](#a-survey-of-structured-pruning-arxiv-version-and-ieee-t-pami-version)\r\n\r\n- [2023 Venues](#2023)\r\n\r\n- [2022 Venues](#2022)\r\n\r\n- [2021 Venues](#2021)\r\n\r\n- [2020 Venues](#2020)\r\n\r\n- [2019 Venues](#2019)\r\n\r\n- [2018 Venues](#2018)\r\n\r\n- [2017 Venues](#2017)\r\n\r\n- [2016 Venues](#2016)\r\n\r\n- [2015 Venues](#2015)\r\n\r\n### Type of Pruning\r\n\r\n| Type | `F` | `W` | `S` | `Other` |\r\n|:----------- |:--------------:|:--------------:|:----------------:|:-----------:|\r\n| Explanation | Filter pruning | Weight pruning | Special Networks | other types |\r\n\r\n### A Survey of Structured Pruning ([arXiv version](https:\u002F\u002Farxiv.org\u002Fabs\u002F2303.00566) and [IEEE T-PAMI version](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F10330640))\r\n\r\nPlease cite our paper if it's helpful:\r\n```\r\n@article{he2024structured,\r\n author={He, Yang and Xiao, Lingao},\r\n journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, \r\n title={Structured Pruning for Deep Convolutional Neural Networks: A Survey}, \r\n year={2024},\r\n volume={46},\r\n number={5},\r\n pages={2900-2919},\r\n doi={10.1109\u002FTPAMI.2023.3334614}}\r\n```\r\n\r\nThe related papers are categorized as below:\r\n![Structured Pruning Taxonomy](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhe-y_Awesome-Pruning_readme_834105cfd147.png)\r\n\r\n### 2023\r\n| Title | Venue | Type | Code |\r\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\r\n| [Revisiting Pruning at Initialization Through the Lens of Ramanujan Graph](https:\u002F\u002Fopenreview.net\u002Fforum?id=uVcDssQff_) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002Framanujan-on-pai)(Releasing) |\r\n| [Unmasking the Lottery Ticket Hypothesis: What's Encoded in a Winning Ticket's Mask?](https:\u002F\u002Fopenreview.net\u002Fforum?id=xSsW2Am-ukZ) | ICLR | `W` | - |\r\n| [Bit-Pruning: A Sparse Multiplication-Less Dot-Product](https:\u002F\u002Fopenreview.net\u002Fforum?id=YUDiZcZTI8) | ICLR | `W` | [Code Deleted](https:\u002F\u002Fgithub.com\u002FDensoITLab\u002Fbitprune) |\r\n| [NTK-SAP: Improving neural network pruning by aligning training dynamics](https:\u002F\u002Fopenreview.net\u002Fforum?id=-5EWhW_4qWP) | ICLR | `W` | - |\r\n| [A Unified Framework for Soft Threshold Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=cCFqcrq0d8) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FYanqi-Chen\u002FLATS) |\r\n| [CrAM: A Compression-Aware Minimizer](https:\u002F\u002Fopenreview.net\u002Fforum?id=_eTZBs-yedr) | ICLR | `W` | - |\r\n| [Trainability Preserving Neural Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=AZFvpnnewr) | ICLR | `F` | - |\r\n| [DFPC: Data flow driven pruning of coupled channels without data](https:\u002F\u002Fopenreview.net\u002Fforum?id=mhnHqRqcjYU) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fdrive.google.com\u002Fdrive\u002Ffolders\u002F18eRYzWnB_6Qq0cYiSzvyOgicqn50g3-m) |\r\n| [TVSPrune - Pruning Non-discriminative filters via Total Variation separability of intermediate representations without fine tuning](https:\u002F\u002Fopenreview.net\u002Fforum?id=sZI1Oj9KBKy) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ftvsprune\u002FTVS_Prune) |\r\n| [HomoDistil: Homotopic Task-Agnostic Distillation of Pre-trained Transformers](https:\u002F\u002Fopenreview.net\u002Fforum?id=D7srTrGhAs) | ICLR | `F` | - |\r\n| [MECTA: Memory-Economic Continual Test-Time Model Adaptation](https:\u002F\u002Fopenreview.net\u002Fforum?id=N92hjSf5NNh) | ICLR | `F` | - |\r\n| [DepthFL : Depthwise Federated Learning for Heterogeneous Clients](https:\u002F\u002Fopenreview.net\u002Fforum?id=pf8RIZTMU58) | ICLR | `F` | - |\r\n| [OTOv2: Automatic, Generic, User-Friendly](https:\u002F\u002Fopenreview.net\u002Fforum?id=7ynoX1ojPMt) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ftianyic\u002Fonly_train_once) |\r\n| [Over-parameterized Model Optimization with Polyak-Lojasiewicz Condition](https:\u002F\u002Fopenreview.net\u002Fforum?id=aBIpZvMdS56) | ICLR | `F` | - |\r\n| [Pruning Deep Neural Networks from a Sparsity Perspective](https:\u002F\u002Fopenreview.net\u002Fforum?id=i-DleYh34BM) | ICLR | `WF` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fdem123456789\u002FPruning-Deep-Neural-Networks-from-a-Sparsity-Perspective) |\r\n| [Holistic Adversarially Robust Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=sAJDi9lD06L) | ICLR | `WF` | - |\r\n| [How I Learned to Stop Worrying and Love Retraining](https:\u002F\u002Fopenreview.net\u002Fforum?id=_nF5imFKQI) | ICLR | `WF` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FZIB-IOL\u002FBIMP) |\r\n| [Symmetric Pruning in Quantum Neural Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=K96AogLDT2K) | ICLR | `S` | - |\r\n| [Rethinking Graph Lottery Tickets: Graph Sparsity Matters](https:\u002F\u002Fopenreview.net\u002Fforum?id=fjh7UGQgOB) | ICLR | `S` | - |\r\n| [Joint Edge-Model Sparse Learning is Provably Efficient for Graph Neural Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=4UldFtZ_CVF) | ICLR | `S` | - |\r\n| [Searching Lottery Tickets in Graph Neural Networks: A Dual Perspective](https:\u002F\u002Fopenreview.net\u002Fforum?id=Dvs-a3aymPe) | ICLR | `S` | - |\r\n| [Diffusion Models for Causal Discovery via Topological Ordering](https:\u002F\u002Fopenreview.net\u002Fforum?id=Idusfje4-Wq) | ICLR | `S` | - |\r\n| [A General Framework For Proving The Equivariant Strong Lottery Ticket Hypothesis](https:\u002F\u002Fopenreview.net\u002Fforum?id=vVJZtlZB9D) | ICLR | `Other` | - |\r\n| [Sparsity May Cry: Let Us Fail (Current) Sparse Neural Networks Together!](https:\u002F\u002Fopenreview.net\u002Fforum?id=J6F3lLg4Kdp) | ICLR | `Other` | - |\r\n| [Minimum Variance Unbiased N:M Sparsity for the Neural Gradients](https:\u002F\u002Fopenreview.net\u002Fforum?id=vuD2xEtxZcj) | ICLR | `Other` | - |\r\n\r\n### 2022\r\n| Title | Venue | Type | Code |\r\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\r\n| [Parameter-Efficient Masking Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=7rcuQ_V2GFg) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyueb17\u002FPEMN) |\r\n| [\"Lossless\" Compression of Deep Neural Networks: A High-dimensional Neural Tangent Kernel Approach](https:\u002F\u002Fopenreview.net\u002Fforum?id=NaW6T93F34m) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FModel-Compression\u002FLossless_Compression) |\r\n| [Losses Can Be Blessings: Routing Self-Supervised Speech Representations Towards Efficient Multilingual and Multitask Speech Processing](https:\u002F\u002Fopenreview.net\u002Fforum?id=2EUJ4e6H4OX) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FGATECH-EIC\u002FS3-Router) |\r\n| [Models Out of Line: A Fourier Lens on Distribution Shift Robustness](https:\u002F\u002Fopenreview.net\u002Fforum?id=YZ-N-sejjwO) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fsarafridov\u002FRobustNets) |\r\n| [Robust Binary Models by Pruning Randomly-initialized Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=5g-h_DILemH) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FIVRL\u002FRobustBinarySubNet) |\r\n| [Rare Gems: Finding Lottery Tickets at Initialization](https:\u002F\u002Fopenreview.net\u002Fforum?id=Jpxd93u2vK-) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fksreenivasan\u002Fpruning_is_enough) |\r\n| [Optimal Brain Compression: A Framework for Accurate Post-Training Quantization and Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=ksVGCOlOEba) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FIST-DASLab\u002FOBC) |\r\n| [Pruning’s Effect on Generalization Through the Lens of Training and Regularization](https:\u002F\u002Fopenreview.net\u002Fforum?id=OrcLKV9sKWp) | NeurIPS | `W` | - |\r\n| [Back Razor: Memory-Efficient Transfer Learning by Self-Sparsified Backpropagation](https:\u002F\u002Fopenreview.net\u002Fforum?id=mTXQIpXPDbh) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FBackRazor_Neurips22) |\r\n| [Analyzing Lottery Ticket Hypothesis from PAC-Bayesian Theory Perspective](https:\u002F\u002Fopenreview.net\u002Fforum?id=fbUybomIuE) | NeurIPS | `W` | - |\r\n| [Sparse Winning Tickets are Data-Efficient Image Recognizers](https:\u002F\u002Fopenreview.net\u002Fforum?id=wfKbtSjHA6F) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FDataEfficientLTH) |\r\n| [Lottery Tickets on a Data Diet: Finding Initializations with Sparse Trainable Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=QLPzCpu756J) | NeurIPS | `W` | - |\r\n| [Weighted Mutual Learning with Diversity-Driven Model Compression](https:\u002F\u002Fopenreview.net\u002Fforum?id=UQJoGBNRX4) | NeurIPS | `F` | - |\r\n| [SInGE: Sparsity via Integrated Gradients Estimation of Neuron Relevance](https:\u002F\u002Fopenreview.net\u002Fforum?id=oQIJsMlyaW_) | NeurIPS | `F` | - |\r\n| [Data-Efficient Structured Pruning via Submodular Optimization](https:\u002F\u002Fopenreview.net\u002Fforum?id=K2QGzyLwpYG) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fmarwash25\u002Fsubpruning) |\r\n| [Structural Pruning via Latency-Saliency Knapsack](https:\u002F\u002Fopenreview.net\u002Fforum?id=cUOR-_VsavA) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FHALP) |\r\n| [Recall Distortion in Neural Network Pruning and the Undecayed Pruning Algorithm](https:\u002F\u002Fopenreview.net\u002Fforum?id=5hgYi4r5MDp) | NeurIPS | `WF` | - |\r\n| [Pruning Neural Networks via Coresets and Convex Geometry: Towards No Assumptions](https:\u002F\u002Fopenreview.net\u002Fforum?id=btpIaJiRx6z) | NeurIPS | `WF` | - |\r\n| [Controlled Sparsity via Constrained Optimization or: How I Learned to Stop Tuning Penalties and Love Constraints](https:\u002F\u002Fopenreview.net\u002Fforum?id=XUvSYc6TqDF) | NeurIPS | `WF` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fgallego-posada\u002Fconstrained_sparsity) |\r\n| [Advancing Model Pruning via Bi-level Optimization](https:\u002F\u002Fopenreview.net\u002Fforum?id=t6O08FxvtBY) | NeurIPS | `WF` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FOPTML-Group\u002FBiP) |\r\n| [Emergence of Hierarchical Layers in a Single Sheet of Self-Organizing Spiking Neurons](https:\u002F\u002Fopenreview.net\u002Fforum?id=cPVuuk1lZb3) | NeurIPS | `S` | - |\r\n| [CryptoGCN: Fast and Scalable Homomorphically Encrypted Graph Convolutional Network Inference](https:\u002F\u002Fopenreview.net\u002Fforum?id=VeQBBm1MmTZ) | NeurIPS | `S` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Franran0523\u002FCryptoGCN)(Releasing) |\r\n| [Transform Once: Efficient Operator Learning in Frequency Domain](https:\u002F\u002Fopenreview.net\u002Fforum?id=B2PpZyAAEgV) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FDiffEqML\u002Fkairos)(Releasing) |\r\n| [Most Activation Functions Can Win the Lottery Without Excessive Depth](https:\u002F\u002Fopenreview.net\u002Fforum?id=NySDKS9SxN) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FRelationalML\u002FLT-existence) |\r\n| [Pruning has a disparate impact on model accuracy](https:\u002F\u002Fopenreview.net\u002Fforum?id=11nMVZK0WYM) | NeurIPS | `Other` | - |\r\n| [Model Preserving Compression for Neural Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=gt-l9Hu2ndd) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fjerry-chee\u002FModelPreserveCompressionNN) |\r\n| [Prune Your Model Before Distill It](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_8) | ECCV | `W` | [PyTorch(Author)](https:\u002F\u002Fhttps:\u002F\u002Fgithub.com\u002Fososos888\u002Fprune-then-distill) |\r\n| [FedLTN: Federated Learning for Sparse and Personalized Lottery Ticket Networks](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19775-8_5) | ECCV | `W` | - |\r\n| [FairGRAPE: Fairness-Aware GRAdient Pruning mEthod for Face Attribute Classification](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19778-9_24) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FBernardo1998\u002FFairGRAPE) |\r\n| [SuperTickets: Drawing Task-Agnostic Lottery Tickets from Supernets via Jointly Architecture Searching and Parameter Pruning](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_40) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FGATECH-EIC\u002FSuperTickets) |\r\n| [Ensemble Knowledge Guided Sub-network Search and Fine-Tuning for Filter Pruning](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_34) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fsseung0703\u002FEKG) |\r\n| [CPrune: Compiler-Informed Model Pruning for Efficient Target-Aware DNN Execution](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20044-1_37) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ftaehokim20\u002FCPrune) |\r\n| [Soft Masking for Cost-Constrained Channel Pruning](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_38) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FSMCP) |\r\n| [Filter Pruning via Feature Discrimination in Deep Neural Networks](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19803-8_15) | ECCV | `F` | - |\r\n| [Disentangled Differentiable Network Pruning](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_20) | ECCV | `F` | - |\r\n| [Interpretations Steered Network Pruning via Amortized Inferred Saliency Maps](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19803-8_17) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FAlii-Ganjj\u002FInterpretationsSteeredPruning) |\r\n| [Bayesian Optimization with Clustering and Rollback for CNN Auto Pruning](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20050-2_29) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ffanhanwei\u002FBOCR) |\r\n| [Multi-granularity Pruning for Model Acceleration on Mobile Devices](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_29) | ECCV | `WF` | - |\r\n| [Exploring Lottery Ticket Hypothesis in Spiking Neural Networks](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19775-8_7) | ECCV | `S` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FIntelligent-Computing-Lab-Yale\u002FExploring-Lottery-Ticket-Hypothesis-in-SNNs) |\r\n| [Towards Ultra Low Latency Spiking Neural Networks for Vision and Sequential Tasks Using Temporal Pruning](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_42) | ECCV | `S` | - |\r\n| [Recent Advances on Neural Network Pruning at Initialization](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F786) | IJCAI | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fmingsun-tse\u002Fsmile-pruning) |\r\n| [FedDUAP: Federated Learning with Dynamic Update and Adaptive Pruning Using Shared Data on the Server](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F385) | IJCAI | `F` | - |\r\n| [On the Channel Pruning using Graph Convolution Network for Convolutional Neural Network Acceleration](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F431) | IJCAI | `F` | - |\r\n| [Pruning-as-Search: Efficient Neural Architecture Search via Channel Pruning and Structural Reparameterization](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F449) | IJCAI | `F` | - |\r\n| [Neural Network Pruning by Cooperative Coevolution](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F667) | IJCAI | `F` | - |\r\n| [SPDY: Accurate Pruning with Speedup Guarantees](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Ffrantar22a.html) | ICML | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FIST-DASLab\u002Fspdy) |\r\n| [Sparse Double Descent: Where Network Pruning Aggravates Overfitting](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fhe22d.html) | ICML | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fhezheug\u002Fsparse-double-descent) |\r\n| [The Combinatorial Brain Surgeon: Pruning Weights That Cancel One Another in Neural Networks](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fyu22f.html) | ICML | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyuxwind\u002FCBS) |\r\n| [Linearity Grafting: Relaxed Neuron Pruning Helps Certifiable Robustness](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fchen22af.html) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FLinearity-Grafting) |\r\n| [Winning the Lottery Ahead of Time: Efficient Early Network Pruning](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Frachwan22a.html) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fjohnrachwan123\u002FEarly-Cropression-via-Gradient-Flow-Preservation) |\r\n| [Topology-Aware Network Pruning using Multi-stage Graph Embedding and Reinforcement Learning](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fyu22e.html) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyusx-swapp\u002FGNN-RL-Model-Compression) |\r\n| [Fast Lossless Neural Compression with Integer-Only Discrete Flows](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fwang22a.html) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fthu-ml\u002FIODF) |\r\n| [DepthShrinker: A New Compression Paradigm Towards Boosting Real-Hardware Efficiency of Compact Neural Networks](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Ffu22c.html) | ICML | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002FDepthShrinker) |\r\n| [PAC-Net: A Model Pruning Approach to Inductive Transfer Learning](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fmyung22a.html) | ICML | `Other` | - |\r\n| [Neural Network Pruning Denoises the Features and Makes Local Connectivity Emerge in Visual Tasks](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fpellegrini22a.html) | ICML | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fphiandark\u002FSiftingFeatures) |\r\n| [Interspace Pruning: Using Adaptive Filter Representations To Improve Training of Sparse CNNs](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FWimmer_Interspace_Pruning_Using_Adaptive_Filter_Representations_To_Improve_Training_of_CVPR_2022_paper.html) | CVPR | `W` | - |\r\n| [Masking Adversarial Damage: Finding Adversarial Saliency for Robust and Sparse Network](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FLee_Masking_Adversarial_Damage_Finding_Adversarial_Saliency_for_Robust_and_Sparse_CVPR_2022_paper.html) | CVPR | `W` | - |\r\n| [When To Prune? A Policy Towards Early Structural Pruning](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FShen_When_To_Prune_A_Policy_Towards_Early_Structural_Pruning_CVPR_2022_paper.html) | CVPR | `F` | - |\r\n| [Fire Together Wire Together: A Dynamic Pruning Approach With Self-Supervised Mask PredictionFire Together Wire Together: A Dynamic Pruning Approach With Self-Supervised Mask Prediction](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FElkerdawy_Fire_Together_Wire_Together_A_Dynamic_Pruning_Approach_With_Self-Supervised_CVPR_2022_paper.html) | CVPR | `F` | - |\r\n| [Revisiting Random Channel Pruning for Neural Network Compression](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FLi_Revisiting_Random_Channel_Pruning_for_Neural_Network_Compression_CVPR_2022_paper.html) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fofsoundof\u002Frandom_channel_pruning)(Releasing) |\r\n| [Learning Bayesian Sparse Networks With Full Experience Replay for Continual Learning](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FYan_Learning_Bayesian_Sparse_Networks_With_Full_Experience_Replay_for_Continual_CVPR_2022_paper.html) | CVPR | `F` | - |\r\n| [DECORE: Deep Compression With Reinforcement Learning](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FAlwani_DECORE_Deep_Compression_With_Reinforcement_Learning_CVPR_2022_paper.html) | CVPR | `F` | - |\r\n| [CHEX: CHannel EXploration for CNN Model Compression](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FHou_CHEX_CHannel_EXploration_for_CNN_Model_Compression_CVPR_2022_paper.html) | CVPR | `F` | - |\r\n| [Compressing Models With Few Samples: Mimicking Then Replacing](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FWang_Compressing_Models_With_Few_Samples_Mimicking_Then_Replacing_CVPR_2022_paper.html) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fcjnjuwhy\u002FMiR)(Releasing) |\r\n| [Contrastive Dual Gating: Learning Sparse Features With Contrastive Learning](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FMeng_Contrastive_Dual_Gating_Learning_Sparse_Features_With_Contrastive_Learning_CVPR_2022_paper.html) | CVPR | `WF` | - |\r\n| [DiSparse: Disentangled Sparsification for Multitask Model Compression](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FSun_DiSparse_Disentangled_Sparsification_for_Multitask_Model_Compression_CVPR_2022_paper.html) | CVPR | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FSHI-Labs\u002FDiSparse-Multitask-Model-Compression) |\r\n| [Learning Pruning-Friendly Networks via Frank-Wolfe: One-Shot, Any-Sparsity, And No Retraining](https:\u002F\u002Fopenreview.net\u002Fforum?id=O1DEtITim__) | ICLR **(Spotlight)** | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FSFW-Once-for-All-Pruning) |\r\n| [On Lottery Tickets and Minimal Task Representations in Deep Reinforcement Learning](https:\u002F\u002Fopenreview.net\u002Fforum?id=Fl3Mg_MZR-) | ICLR **(Spotlight)** | `W` | - |\r\n| [An Operator Theoretic View On Pruning Deep Neural Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=pWBNOgdeURp) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fwilliam-redman\u002FKoopman_pruning) |\r\n| [Effective Model Sparsification by Scheduled Grow-and-Prune Methods](https:\u002F\u002Fopenreview.net\u002Fforum?id=xa6otUDdP2W) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fboone891214\u002FGaP) |\r\n| [Signing the Supermask: Keep, Hide, Invert](https:\u002F\u002Fopenreview.net\u002Fforum?id=e0jtGTfPihs) | ICLR | `W` | - |\r\n| [How many degrees of freedom do we need to train deep networks: a loss landscape perspective](https:\u002F\u002Fopenreview.net\u002Fforum?id=ChMLTGRjFcU) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fganguli-lab\u002Fdegrees-of-freedom) |\r\n| [Dual Lottery Ticket Hypothesis](https:\u002F\u002Fopenreview.net\u002Fforum?id=fOsN52jn25l) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyueb17\u002FDLTH) |\r\n| [Peek-a-Boo: What (More) is Disguised in a Randomly Weighted Neural Network, and How to Find It Efficiently](https:\u002F\u002Fopenreview.net\u002Fforum?id=moHCzz6D5H3) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FPeek-a-Boo) |\r\n| [Sparsity Winning Twice: Better Robust Generalization from More Efficient Training](https:\u002F\u002Fopenreview.net\u002Fforum?id=SYuJXrXq8tw) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FSparsity-Win-Robust-Generalization) |\r\n| [SOSP: Efficiently Capturing Global Correlations by Second-Order Structured Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=t5EmXZ3ZLR) | ICLR **(Spotlight)** | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fboschresearch\u002Fsosp)(Releasing) |\r\n| [Pixelated Butterfly: Simple and Efficient Sparse training for Neural Network Models](https:\u002F\u002Fopenreview.net\u002Fforum?id=Nfl-iXa-y7R) | ICLR **(Spotlight)** | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FHazyResearch\u002Fpixelfly) |\r\n| [Revisit Kernel Pruning with Lottery Regulated Grouped Convolutions](https:\u002F\u002Fopenreview.net\u002Fforum?id=LdEhiMG9WLO) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FchoH\u002Flottery_regulated_grouped_kernel_pruning) |\r\n| [Plant 'n' Seek: Can You Find the Winning Ticket?](https:\u002F\u002Fopenreview.net\u002Fforum?id=9n9c8sf0xm) | ICLR | `F` | [PyTorch(Author)](http:\u002F\u002Fwww.github.com\u002FRelationalML\u002FPlantNSeek) |\r\n| [Proving the Lottery Ticket Hypothesis for Convolutional Neural Networks](https:\u002F\u002Fopenreview.net\u002Fforum?id=Vjki79-619-) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FArthurWalraven\u002Fcnnslth) |\r\n| [On the Existence of Universal Lottery Tickets](https:\u002F\u002Fopenreview.net\u002Fforum?id=SYB4WrJql1n) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FRelationalML\u002FUniversalLT) |\r\n| [Training Structured Neural Networks Through Manifold Identification and Variance Reduction](https:\u002F\u002Fopenreview.net\u002Fforum?id=mdUYT5QV0O) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fwww.github.com\u002Fzihsyuan1214\u002Frmda) |\r\n| [Learning Efficient Image Super-Resolution Networks via Structure-Regularized Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=AjGC97Aofee) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FMingSun-Tse\u002FSRP) |\r\n| [Prospect Pruning: Finding Trainable Weights at Initialization using Meta-Gradients](https:\u002F\u002Fopenreview.net\u002Fforum?id=AIgn9uwfcD1) | ICLR | `WF` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fmil-ad\u002Fprospr) |\r\n| [The Unreasonable Effectiveness of Random Pruning: Return of the Most Naive Baseline for Sparse Training](https:\u002F\u002Fopenreview.net\u002Fforum?id=VBZJ_3tz-t) | ICLR | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FRandom_Pruning) |\r\n| [Prune and Tune Ensembles: Low-Cost Ensemble Learning with Sparse Independent Subnetworks](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F20842) | AAAI | `W` | - |\r\n| [Prior Gradient Mask Guided Pruning-Aware Fine-Tuning](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F19888) | AAAI | `F` | - |\r\n| [Convolutional Neural Network Compression through Generalized Kronecker Product Decomposition](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F19958) | AAAI | `Other` | - |\r\n\r\n### 2021\r\n| Title | Venue | Type | Code |\r\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\r\n| [Validating the Lottery Ticket Hypothesis with Inertial Manifold Theory](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Ffdc42b6b0ee16a2f866281508ef56730-Abstract.html) | NeurIPS | `W` | - |\r\n| [The Elastic Lottery Ticket Hypothesis](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fdfccdb8b1cc7e4dab6d33db0fef12b88-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FElasticLTH) |\r\n| [Sanity Checks for Lottery Tickets: Does Your Winning Ticket Really Win the Jackpot?](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F6a130f1dc6f0c829f874e92e5458dced-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fboone891214\u002Fsanity-check-LTH) |\r\n| [Why Lottery Ticket Wins? A Theoretical Perspective of Sample Complexity on Sparse Neural Networks](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F15f99f2165aa8c86c9dface16fefd281-Abstract.html) | NeurIPS | `W` | - |\r\n| [You are caught stealing my winning lottery ticket! Making a lottery ticket claim its ownership](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F23e582ad8087f2c03a5a31c125123f9a-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FNO-stealing-LTH) |\r\n| [Pruning Randomly Initialized Neural Networks with Iterative Randomization](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F23e582ad8087f2c03a5a31c125123f9a-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fdchiji-ntt\u002Fiterand) |\r\n| [Sparse Training via Boosting Pruning Plasticity with Neuroregeneration](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F5227b6aaf294f5f027273aebf16015f2-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FGraNet) |\r\n| [AC\u002FDC: Alternating Compressed\u002FDeCompressed Training of Deep Neural Networks](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F48000647b315f6f00f913caa757a70b3-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FIST-DASLab\u002FACDC) |\r\n| [A Winning Hand: Compressing Deep Networks Can Improve Out-of-Distribution Robustness](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F0607f4c705595b911a4f3e7a127b44e0-Abstract.html) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FRobustBench\u002Frobustbench) |\r\n| [Rethinking the Pruning Criteria for Convolutional Neural Network](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F87ae6fb631f7c8a627e8e28785d9992d-Abstract.html) | NeurIPS | `F` | - |\r\n| [Only Train Once: A One-Shot Neural Network Training And Pruning Framework](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fa376033f78e144f494bfc743c0be3330-Abstract.html) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ftianyic\u002Fonlytrainonce) |\r\n| [CHIP: CHannel Independence-based Pruning for Compact Neural Networks](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fce6babd060aa46c61a5777902cca78af-Abstract.html) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FEclipsess\u002FCHIP_NeurIPS2021) |\r\n| [RED : Looking for Redundancies for Data-FreeStructured Compression of Deep Neural Networks](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fae5e3ce40e0404a45ecacaaf05e5f735-Abstract.html) | NeurIPS | `F` | - |\r\n| [Compressing Neural Networks: Towards Determining the Optimal Layer-wise Decomposition](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F2adcfc3929e7c03fac3100d3ad51da26-Abstract.html) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Flucaslie\u002Ftorchprune) |\r\n| [Sparse Flows: Pruning Continuous-depth Models](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fbf1b2f4b901c21a1d8645018ea9aeb05-Abstract.html) | NeurIPS | `WF` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Flucaslie\u002Ftorchprune) |\r\n| [Scaling Up Exact Neural Network Compression by ReLU Stability](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fe35d7a5768c4b85b4780384d55dc3620-Abstract.html) | NeurIPS | `S` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyuxwind\u002FExactCompression) |\r\n| [Discriminator in GAN Compression: A Generator-discriminator Cooperative Compression Scheme](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Feffc299a1addb07e7089f9b269c31f2f-Abstract.html) | NeurIPS | `S` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FSJLeo\u002FGCC) |\r\n| [Heavy Tails in SGD and Compressibility of Overparametrized Neural Networks](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Ff5c3dd7514bf620a1b85450d2ae374b1-Abstract.html) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fmbarsbey\u002Fsgd_comp_gen) |\r\n| [ResRep: Lossless CNN Pruning via Decoupling Remembering and Forgetting](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FDing_ResRep_Lossless_CNN_Pruning_via_Decoupling_Remembering_and_Forgetting_ICCV_2021_paper.html) | ICCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FDingXiaoH\u002FResRep) |\r\n| [Achieving on-Mobile Real-Time Super-Resolution with Neural Architecture and Pruning Search](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FZhan_Achieving_On-Mobile_Real-Time_Super-Resolution_With_Neural_Architecture_and_Pruning_Search_ICCV_2021_paper.html) | ICCV | `F` | - |\r\n| [GDP: Stabilized Neural Network Pruning via Gates with Differentiable Polarization](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FGuo_GDP_Stabilized_Neural_Network_Pruning_via_Gates_With_Differentiable_Polarization_ICCV_2021_paper.html) | ICCV | `F` | - |\r\n| [Auto Graph Encoder-Decoder for Neural Network Pruning](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FYu_Auto_Graph_Encoder-Decoder_for_Neural_Network_Pruning_ICCV_2021_paper.html) | ICCV | `F` | - |\r\n| [Exploration and Estimation for Model Compression](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F5227b6aaf294f5f027273aebf16015f2-Abstract.html) | ICCV | `F` | - |\r\n| [Sub-Bit Neural Networks: Learning To Compress and Accelerate Binary Neural Networks](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FWang_Sub-Bit_Neural_Networks_Learning_To_Compress_and_Accelerate_Binary_Neural_ICCV_2021_paper.html) | ICCV | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyikaiw\u002FSNN) |\r\n| [On the Predictability of Pruning Across Scales](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.10621) | ICML | `W` | - |\r\n| [A Probabilistic Approach to Neural Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.10065) | ICML | `F` | - |\r\n| [Accelerate CNNs from Three Dimensions: A Comprehensive Pruning Framework](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.04879) | ICML | `F` | - |\r\n| [Group Fisher Pruning for Practical Network Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.00708) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fjshilong\u002FFisherPruning) |\r\n| [Towards Compact CNNs via Collaborative Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.11228) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fliuguoyou\u002FTowards-Compact-CNNs-via-Collaborative-Compression) |\r\n| [Permute, Quantize, and Fine-tune: Efficient Compression of Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.15703) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fuber-research\u002Fpermute-quantize-finetune) |\r\n| [NPAS: A Compiler-aware Framework of Unified Network Pruning andArchitecture Search for Beyond Real-Time Mobile Acceleration](https:\u002F\u002Farxiv.org\u002Fabs\u002F2012.00596) | CVPR | `F` | - |\r\n| [Network Pruning via Performance Maximization](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2021\u002Fhtml\u002FGao_Network_Pruning_via_Performance_Maximization_CVPR_2021_paper.html) | CVPR | `F` | - |\r\n| [Convolutional Neural Network Pruning with Structural Redundancy Reduction](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.03438) | CVPR | `F` | - |\r\n| [Manifold Regularized Dynamic Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.05861) | CVPR | `F` | - |\r\n| [Joint-DetNAS: Upgrade Your Detector with NAS, Pruning and Dynamic Distillation](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.12971) | CVPR | `FO` | - |\r\n| [Content-Aware GAN Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.02244) | CVPR | `S` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Flychenyoko\u002Fcontent-aware-gan-compression) |\r\n| [Multi-Prize Lottery Ticket Hypothesis: Finding Accurate Binary Neural Networks by Pruning A Randomly Weighted Network](https:\u002F\u002Fopenreview.net\u002Fforum?id=U_mat0b9iv) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fchrundle\u002Fbiprop) |\r\n| [Layer-adaptive Sparsity for the Magnitude-based Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=H6ATjJ0TKdf) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fjaeho-lee\u002Flayer-adaptive-sparsity) |\r\n| [Pruning Neural Networks at Initialization: Why Are We Missing the Mark?](https:\u002F\u002Fopenreview.net\u002Fforum?id=Ig-VyQc-MLK) | ICLR | `W` | - |\r\n| [Robust Pruning at Initialization](https:\u002F\u002Fopenreview.net\u002Fforum?id=vXj_ucZQ4hA) | ICLR | `W` | - |\r\n| [A Gradient Flow Framework For Analyzing Network Pruning](https:\u002F\u002Fopenreview.net\u002Fforum?id=rumv7QmLUue) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FEkdeepSLubana\u002Fflowandprune) |\r\n| [Neural Pruning via Growing Regularization](https:\u002F\u002Fopenreview.net\u002Fforum?id=o966_Is_nPA) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FMingSun-Tse\u002FRegularization-Pruning) |\r\n| [ChipNet: Budget-Aware Pruning with Heaviside Continuous Approximations](https:\u002F\u002Fopenreview.net\u002Fforum?id=xCxXwTzx4L1) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FtransmuteAI\u002FChipNet) |\r\n| [Network Pruning That Matters: A Case Study on Retraining Variants](https:\u002F\u002Fopenreview.net\u002Fforum?id=Cb54AMqHQFP) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Flehduong\u002FNPTM) |\r\n\r\n### 2020\r\n\r\n| Title | Venue | Type | Code |\r\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\r\n| [Optimal Lottery Tickets via Subset Sum: Logarithmic Over-Parameterization is Sufficient](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2020\u002Fhash\u002F1b742ae215adf18b75449c6e272fd92d-Abstract.html) | NeurIPS | `W` | - |\r\n| [Winning the Lottery with Continuous Sparsification](https:\u002F\u002Farxiv.org\u002Fabs\u002F1912.04427v4) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Flolemacs\u002Fcontinuous-sparsification) |\r\n| [HYDRA: Pruning Adversarially Robust Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.10509) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Finspire-group\u002Fhydra) |\r\n| [Logarithmic Pruning is All You Need](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.12156) | NeurIPS | `W` | - |\r\n| [Directional Pruning of Deep Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.09358) | NeurIPS | `W` | - |\r\n| [Movement Pruning: Adaptive Sparsity by Fine-Tuning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.07683) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fblock_movement_pruning) |\r\n| [Sanity-Checking Pruning Methods: Random Tickets can Win the Jackpot](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.11094) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FJingtongSu\u002Fsanity-checking-pruning) |\r\n| [Neuron Merging: Compensating for Pruned Neurons](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.13160) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ffriendshipkim\u002Fneuron-merging) |\r\n| [Neuron-level Structured Pruning using Polarization Regularizer](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2020\u002Ffile\u002F703957b6dd9e3a7980e040bee50ded65-Paper.pdf) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fpolarizationpruning\u002FPolarizationPruning) |\r\n| [SCOP: Scientific Control for Reliable Neural Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.10732) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyehuitang\u002FPruning\u002Ftree\u002Fmaster\u002FSCOP_NeurIPS2020) |\r\n| [Storage Efficient and Dynamic Flexible Runtime Channel Pruning via Deep Reinforcement Learning](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2020\u002Fhash\u002Fa914ecef9c12ffdb9bede64bb703d877-Abstract.html) | NeurIPS | `F` | - |\r\n| [The Generalization-Stability Tradeoff In Neural Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.03728) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fbbartoldson\u002FGeneralizationStabilityTradeoff) |\r\n| [Greedy Optimization Provably Wins the Lottery: Logarithmic Number of Winning Tickets is Enough](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2020\u002Fhash\u002Fbe23c41621390a448779ee72409e5f49-Abstract.html) | NeurIPS | `WF` | - |\r\n| [Pruning Filter in Filter](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.14410) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Ffxmeng\u002FPruning-Filter-in-Filter) |\r\n| [Position-based Scaled Gradient for Model Quantization and Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11035) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FJangho-Kim\u002FPSG-pytorch) |\r\n| [Bayesian Bits: Unifying Quantization and Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.07093) | NeurIPS | `Other` | - |\r\n| [Pruning neural networks without any data by iteratively conserving synaptic flow](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.05467) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fganguli-lab\u002FSynaptic-Flow) |\r\n| [Meta-Learning with Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.03219) | ECCV | `W` | - |\r\n| [Accelerating CNN Training by Pruning Activation Gradients](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.00173) | ECCV | `W` | - |\r\n| [EagleEye: Fast Sub-net Evaluation for Efficient Neural Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.02491) | ECCV **(Oral)** | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fanonymous47823493\u002FEagleEye) |\r\n| [DSA: More Efficient Budgeted Pruning via Differentiable Sparsity Allocation](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.02164) | ECCV | `F` | - |\r\n| [DHP: Differentiable Meta Pruning via HyperNetworks](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.13683) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fofsoundof\u002Fdhp) |\r\n| [DA-NAS: Data Adapted Pruning for Efficient Neural Architecture Search](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.12563) S | ECCV | `Other` | - |\r\n| [Differentiable Joint Pruning and Quantization for Hardware Efficiency](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.10463) | ECCV | `Other` | - |\r\n| [Channel Pruning via Automatic Structure Search](https:\u002F\u002Farxiv.org\u002Fabs\u002F2001.08565) | IJCAI | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Flmbxmu\u002FABCPruner) |\r\n| [Adversarial Neural Pruning with Latent Vulnerability Suppression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.04355) | ICML | `W` | - |\r\n| [Proving the Lottery Ticket Hypothesis: Pruning is All You Need](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.00585) | ICML | `W` | - |\r\n| [Network Pruning by Greedy Subnetwork Selection](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.01794) | ICML | `F` | - |\r\n| [Operation-Aware Soft Channel Pruning using Differentiable Masks](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.03938) | ICML | `F` | - |\r\n| [DropNet: Reducing Neural Network Complexity via Iterative Pruning](https:\u002F\u002Fproceedings.mlr.press\u002Fv119\u002Ftan20a.html) | ICML | `F` | - |\r\n| [Soft Threshold Weight Reparameterization for Learnable Sparsity](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.03231) | ICML | `WF` | [Pytorch(Author)](https:\u002F\u002Fgithub.com\u002FRAIVNLab\u002FSTR) |\r\n| [Structured Compression by Weight Encryption for Unstructured Pruning and Quantization](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.10138) | CVPR | `W` | - |\r\n| [Automatic Neural Network Compression by Sparsity-Quantization Joint Learning: A Constrained Optimization-Based Approach](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fpapers\u002FYang_Automatic_Neural_Network_Compression_by_Sparsity-Quantization_Joint_Learning_A_Constrained_CVPR_2020_paper.pdf) | CVPR | `W` | - |\r\n| [Towards Efficient Model Compression via Learned Global Ranking](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.12368) | CVPR **(Oral)** | `F` | [Pytorch(Author)](https:\u002F\u002Fgithub.com\u002Fcmu-enyac\u002FLeGR) |\r\n| [HRank: Filter Pruning using High-Rank Feature Map](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.10179) | CVPR **(Oral)** | `F` | [Pytorch(Author)](https:\u002F\u002Fgithub.com\u002Flmbxmu\u002FHRank) |\r\n| [Neural Network Pruning with Residual-Connections and Limited-Data](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.08114) | CVPR **(Oral)** | `F` | - |\r\n| [DMCP: Differentiable Markov Channel Pruning for Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.03354) | CVPR **(Oral)** | `F` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Fzx55\u002Fdmcp) |\r\n| [Group Sparsity: The Hinge Between Filter Pruning and Decomposition for Network Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.08935) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fofsoundof\u002Fgroup_sparsity) |\r\n| [Few Sample Knowledge Distillation for Efficient Network Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.01839) | CVPR | `F` | - |\r\n| [Discrete Model Compression With Resource Constraint for Deep Neural Networks](http:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fhtml\u002FGao_Discrete_Model_Compression_With_Resource_Constraint_for_Deep_Neural_Networks_CVPR_2020_paper.html) | CVPR | `F` | - |\r\n| [Learning Filter Pruning Criteria for Deep Convolutional Neural Networks Acceleration](http:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fhtml\u002FHe_Learning_Filter_Pruning_Criteria_for_Deep_Convolutional_Neural_Networks_Acceleration_CVPR_2020_paper.html) | CVPR | `F` | - |\r\n| [APQ: Joint Search for Network Architecture, Pruning and Quantization Policy](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.08509) | CVPR | `F` | - |\r\n| [Multi-Dimensional Pruning: A Unified Framework for Model Compression](http:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fhtml\u002FGuo_Multi-Dimensional_Pruning_A_Unified_Framework_for_Model_Compression_CVPR_2020_paper.html) | CVPR **(Oral)** | `WF` | - |\r\n| [A Signal Propagation Perspective for Pruning Neural Networks at Initialization](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.06307) | ICLR **(Spotlight)** | `W` | - |\r\n| [ProxSGD: Training Structured Neural Networks under Regularization and Constraints](https:\u002F\u002Fopenreview.net\u002Fforum?id=HygpthEtvr) | ICLR | `W` | [TF+PT(Author)](https:\u002F\u002Fgithub.com\u002Foptyang\u002Fproxsgd) |\r\n| [One-Shot Pruning of Recurrent Neural Networks by Jacobian Spectrum Evaluation](https:\u002F\u002Farxiv.org\u002Fabs\u002F1912.00120) | ICLR | `W` | - |\r\n| [Lookahead: A Far-sighted Alternative of Magnitude-based Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.04809) | ICLR | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Falinlab\u002Flookahead_pruning) |\r\n| [Data-Independent Neural Pruning via Coresets](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.04018) | ICLR | `W` | - |\r\n| [Provable Filter Pruning for Efficient Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.07412) | ICLR | `F` | - |\r\n| [Dynamic Model Pruning with Feedback](https:\u002F\u002Fopenreview.net\u002Fforum?id=SJem8lSFwB) | ICLR | `WF` | - |\r\n| [Comparing Rewinding and Fine-tuning in Neural Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.02389) | ICLR **(Oral)** | `WF` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Flottery-ticket\u002Frewinding-iclr20-public) |\r\n| [AutoCompress: An Automatic DNN Structured Pruning Framework for Ultra-High Compression Rates](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.03141) | AAAI | `F` | - |\r\n| [Reborn filters: Pruning convolutional neural networks with limited data](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F6058) | AAAI | `F` | - |\r\n| [DARB: A Density-Aware Regular-Block Pruning for Deep Neural Networks](http:\u002F\u002Farxiv.org\u002Fabs\u002F1911.08020) | AAAI | `Other` | - |\r\n| [Pruning from Scratch](http:\u002F\u002Farxiv.org\u002Fabs\u002F1909.12579) | AAAI | `Other` | - |\r\n\r\n### 2019\r\n\r\n| Title | Venue | Type | Code |\r\n|:-------|:--------:|:-------:|:-------:|\r\n| [Deconstructing Lottery Tickets: Zeros, Signs, and the Supermask](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.01067) | NeurIPS | `W` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Fuber-research\u002Fdeconstructing-lottery-tickets) |\r\n| [One ticket to win them all: generalizing lottery ticket initializations across datasets and optimizers](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.02773) | NeurIPS | `W` | - |\r\n| [Global Sparse Momentum SGD for Pruning Very Deep Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.12778) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FDingXiaoH\u002FGSM-SGD) |\r\n| [AutoPrune: Automatic Network Pruning by Regularizing Auxiliary Parameters](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F9521-autoprune-automatic-network-pruning-by-regularizing-auxiliary-parameters) | NeurIPS | `W` | - |\r\n| [Network Pruning via Transformable Architecture Search](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.09717) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FD-X-Y\u002FNAS-Projects) |\r\n| [Gate Decorator: Global Filter Pruning Method for Accelerating Deep Convolutional Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.08174) | NeurIPS | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyouzhonghui\u002Fgate-decorator-pruning) |\r\n| [Model Compression with Adversarial Robustness: A Unified Optimization Framework](https:\u002F\u002Farxiv.org\u002Fabs\u002F1902.03538) | NeurIPS | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FTAMU-VITA\u002FATMC) |\r\n| [Adversarial Robustness vs Model Compression, or Both?](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.12561) | ICCV | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fyeshaokai\u002FRobustness-Aware-Pruning-ADMM) |\r\n| [MetaPruning: Meta Learning for Automatic Neural Network Channel Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.10258) | ICCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fliuzechun\u002FMetaPruning) |\r\n| [Accelerate CNN via Recursive Bayesian Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.00353) | ICCV | `F` | - |\r\n| [Learning Filter Basis for Convolutional Neural Network Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.08932) | ICCV | `Other` | - |\r\n| [Co-Evolutionary Compression for Unpaired Image Translation](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.10804) | ICCV | `S` | - |\r\n| [COP: Customized Deep Model Compression via Regularized Correlation-Based Filter-Level Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.10337) | IJCAI | `F` | [Tensorflow(Author)](https:\u002F\u002Fgithub.com\u002FZJULearning\u002FCOP) |\r\n| [Filter Pruning via Geometric Median for Deep Convolutional Neural Networks Acceleration](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.00250) | CVPR **(Oral)** | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Ffilter-pruning-geometric-median) |\r\n| [Towards Optimal Structured CNN Pruning via Generative Adversarial Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.09291) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FShaohuiLin\u002FGAL) |\r\n| [Centripetal SGD for Pruning Very Deep Convolutional Networks with Complicated Structure](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.03837) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FShawnDing1994\u002FCentripetal-SGD) |\r\n| [On Implicit Filter Level Sparsity in Convolutional Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.12495), [Extension1](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.04967), [Extension2](https:\u002F\u002Fopenreview.net\u002Fforum?id=rylVvNS3hE) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fmehtadushy\u002FSelecSLS-Pytorch) |\r\n| [Structured Pruning of Neural Networks with Budget-Aware Regularization](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.09332) | CVPR | `F` | - |\r\n| [Importance Estimation for Neural Network Pruning](http:\u002F\u002Fjankautz.com\u002Fpublications\u002FImportance4NNPruning_CVPR19.pdf) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FTaylor_pruning) |\r\n| [OICSR: Out-In-Channel Sparsity Regularization for Compact Deep Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.11664) | CVPR | `F` | - |\r\n| [Variational Convolutional Neural Network Pruning](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2019\u002Fhtml\u002FZhao_Variational_Convolutional_Neural_Network_Pruning_CVPR_2019_paper.html) | CVPR | `F` | - |\r\n| [Partial Order Pruning: for Best Speed\u002FAccuracy Trade-off in Neural Architecture Search](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.03777) | CVPR | `Other` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Flixincn2015\u002FPartial-Order-Pruning) |\r\n| [Collaborative Channel Pruning for Deep Networks](http:\u002F\u002Fproceedings.mlr.press\u002Fv97\u002Fpeng19c.html) | ICML | `F` | - |\r\n| [Approximated Oracle Filter Pruning for Destructive CNN Width Optimization github](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.04748) | ICML | `F` | - |\r\n| [EigenDamage: Structured Pruning in the Kronecker-Factored Eigenbasis](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.05934) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Falecwangcq\u002FEigenDamage-Pytorch) |\r\n| [The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1803.03635) | ICLR **(Best)** | `W` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Fgoogle-research\u002Flottery-ticket-hypothesis) |\r\n| [SNIP: Single-shot Network Pruning based on Connection Sensitivity](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.02340) | ICLR | `W` | [TensorFLow(Author)](https:\u002F\u002Fgithub.com\u002Fnamhoonlee\u002Fsnip-public) |\r\n| [Dynamic Channel Pruning: Feature Boosting and Suppression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.05331) | ICLR | `F` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Fdeep-fry\u002Fmayo) |\r\n| [Rethinking the Value of Network Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.05270) | ICLR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FEric-mingjie\u002Frethinking-network-pruning) |\r\n| [Dynamic Sparse Graph for Efficient Deep Learning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.00859) | ICLR | `F` | [CUDA(3rd)](https:\u002F\u002Fgithub.com\u002Fmtcrawshaw\u002Fdynamic-sparse-graph) |\r\n\r\n### 2018\r\n| Title | Venue | Type | Code |\r\n|:-------|:--------:|:-------:|:-------:|\r\n| [Frequency-Domain Dynamic Pruning for Convolutional Neural Networks](https:\u002F\u002Fpapers.NeurIPS.cc\u002Fpaper\u002F7382-frequency-domain-dynamic-pruning-for-convolutional-neural-networks.pdf) | NeurIPS | `W` | - |\r\n| [Discrimination-aware Channel Pruning for Deep Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.11809) | NeurIPS | `F` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002FSCUT-AILab\u002FDCP) |\r\n| [Learning Sparse Neural Networks via Sensitivity-Driven Regularization](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1810.11764.pdf) | NeurIPS | `WF` | - |\r\n| [Constraint-Aware Deep Neural Network Compression](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_ECCV_2018\u002Fhtml\u002FChangan_Chen_Constraints_Matter_in_ECCV_2018_paper.html) | ECCV | `W` | [SkimCaffe(Author)](https:\u002F\u002Fgithub.com\u002FChanganVR\u002FConstraintAwareCompression) |\r\n| [A Systematic DNN Weight Pruning Framework using Alternating Direction Method of Multipliers](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.03294) | ECCV | `W` | [Caffe(Author)](https:\u002F\u002Fgithub.com\u002FKaiqiZhang\u002Fadmm-pruning) |\r\n| [Amc: Automl for model compression and acceleration on mobile devices](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.03494) | ECCV | `F` | [TensorFlow(3rd)](https:\u002F\u002Fgithub.com\u002FTencent\u002FPocketFlow#channel-pruning) |\r\n| [Data-Driven Sparse Structure Selection for Deep Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.01213) | ECCV | `F` | [MXNet(Author)](https:\u002F\u002Fgithub.com\u002FTuSimple\u002Fsparse-structure-selection) |\r\n| [Coreset-Based Neural Network Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09810) | ECCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fmetro-smiles\u002FCNN_Compression) |\r\n| [Soft Filter Pruning for Accelerating Deep Convolutional Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1808.06866) | IJCAI | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fsoft-filter-pruning) |\r\n| [Accelerating Convolutional Networks via Global & Dynamic Filter Pruning](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2018\u002F0336.pdf) | IJCAI | `F` | - |\r\n| [Weightless: Lossy weight encoding for deep neural network compression](https:\u002F\u002Fproceedings.mlr.press\u002Fv80\u002Freagan18a.html) | ICML | `W` | - |\r\n| [Compressing Neural Networks using the Variational Information Bottleneck](https:\u002F\u002Fproceedings.mlr.press\u002Fv80\u002Fdai18d.html) | ICML | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fzhuchen03\u002FVIBNet) |\r\n| [Deep k-Means: Re-Training and Parameter Sharing with Harder Cluster Assignments for Compressing Deep Convolutions](https:\u002F\u002Fproceedings.mlr.press\u002Fv80\u002Fwu18h.html) | ICML | `Other` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FDeep-K-Means-pytorch) |\r\n| [CLIP-Q: Deep Network Compression Learning by In-Parallel Pruning-Quantization](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_cvpr_2018\u002Fhtml\u002FTung_CLIP-Q_Deep_Network_CVPR_2018_paper.html) | CVPR | `W` | - |\r\n| [“Learning-Compression” Algorithms for Neural Net Pruning](http:\u002F\u002Ffaculty.ucmerced.edu\u002Fmcarreira-perpinan\u002Fpapers\u002Fcvpr18.pdf) | CVPR | `W` | - |\r\n| [PackNet: Adding Multiple Tasks to a Single Network by Iterative Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.05769) | CVPR | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Farunmallya\u002Fpacknet) |\r\n| [NISP: Pruning Networks using Neuron Importance Score Propagation](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.05908) | CVPR | `F` | - |\r\n| [To prune, or not to prune: exploring the efficacy of pruning for model compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1710.01878) | ICLR | `W` | - |\r\n| [Rethinking the Smaller-Norm-Less-Informative Assumption in Channel Pruning of Convolution Layers](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.00124) | ICLR | `F` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002Fbobye\u002Fbatchnorm_prune), [PyTorch(3rd)](https:\u002F\u002Fgithub.com\u002Fjack-willturner\u002Fbatchnorm-pruning) |\r\n\r\n\r\n### 2017\r\n| Title | Venue | Type | Code |\r\n|:-------|:--------:|:-------:|:-------:|\r\n| [Net-Trim: Convex Pruning of Deep Neural Networks with Performance Guarantee](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05162) | NeurIPS | `W` | [TensorFlow(Author)](https:\u002F\u002Fgithub.com\u002FDNNToolBox\u002FNet-Trim-v1) |\r\n| [Learning to Prune Deep Neural Networks via Layer-wise Optimal Brain Surgeon](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.07565) | NeurIPS | `W` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002Fcsyhhu\u002FL-OBS) |\r\n| [Runtime Neural Pruning](https:\u002F\u002Fpapers.NeurIPS.cc\u002Fpaper\u002F6813-runtime-neural-pruning) | NeurIPS | `F` | - |\r\n| [Structured Bayesian Pruning via Log-Normal Multiplicative Noise](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2017\u002Fhash\u002Fdab49080d80c724aad5ebf158d63df41-Abstract.html) | NeurIPS | `F` | - |\r\n| [Bayesian Compression for Deep Learning](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2017\u002Fhash\u002F69d1fc78dbda242c43ad6590368912d4-Abstract.html) | NeurIPS | `F` | - |\r\n| [ThiNet: A Filter Level Pruning Method for Deep Neural Network Compression](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.06342) | ICCV | `F` | [Caffe(Author)](https:\u002F\u002Fgithub.com\u002FRoll920\u002FThiNet), [PyTorch(3rd)](https:\u002F\u002Fgithub.com\u002Ftranorrepository\u002Freprod-thinet) |\r\n| [Channel pruning for accelerating very deep neural networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.06168) | ICCV | `F` | [Caffe(Author)](https:\u002F\u002Fgithub.com\u002Fyihui-he\u002Fchannel-pruning) |\r\n| [Learning Efficient Convolutional Networks Through Network Slimming](https:\u002F\u002Farxiv.org\u002Fabs\u002F1708.06519) | ICCV | `F` | [PyTorch(Author)](https:\u002F\u002Fgithub.com\u002FEric-mingjie\u002Fnetwork-slimming) |\r\n| [Variational Dropout Sparsifies Deep Neural Networks](http:\u002F\u002Farxiv.org\u002Fabs\u002F1701.05369) | ICML | `W` | - |\r\n| [Combined Group and Exclusive Sparsity for Deep Neural Networks](https:\u002F\u002Fproceedings.mlr.press\u002Fv70\u002Fyoon17a.html) | ICML | `WF` | - |\r\n| [Designing Energy-Efficient Convolutional Neural Networks using Energy-Aware Pruning](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05128) | CVPR | `W` | - |\r\n| [Pruning Filters for Efficient ConvNets](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.08710) | ICLR | `F` | [PyTorch(3rd)](https:\u002F\u002Fgithub.com\u002FEric-mingjie\u002Frethinking-network-pruning\u002Ftree\u002Fmaster\u002Fimagenet\u002Fl1-norm-pruning) |\r\n| [Pruning Convolutional Neural Networks for Resource Efficient Inference](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.06440) | ICLR | `F` | [TensorFlow(3rd)](https:\u002F\u002Fgithub.com\u002FTencent\u002FPocketFlow#channel-pruning) |\r\n\r\n\r\n### 2016\r\n| Title | Venue | Type | Code |\r\n|:-------|:--------:|:-------:|:-------:|\r\n| [Dynamic Network Surgery for Efficient DNNs](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.04493) | NeurIPS | `W` | [Caffe(Author)](https:\u002F\u002Fgithub.com\u002Fyiwenguo\u002FDynamic-Network-Surgery) |\r\n| [Learning the Number of Neurons in Deep Networks](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2016\u002Fhash\u002F6e7d2da6d3953058db75714ac400b584-Abstract.html) | NeurIPS | `F` | - |\r\n| [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https:\u002F\u002Farxiv.org\u002Fabs\u002F1510.00149) | ICLR **(Best)** | `W` | [Caffe(Author)](https:\u002F\u002Fgithub.com\u002Fsonghan\u002FDeep-Compression-AlexNet) |\r\n\r\n\r\n### 2015\r\n\r\n| Title | Venue | Type | Code |\r\n|:-------|:--------:|:-------:|:-------:|\r\n| [Learning both Weights and Connections for Efficient Neural Networks](https:\u002F\u002Farxiv.org\u002Fabs\u002F1506.02626) | NeurIPS | `W` | [PyTorch(3rd)](https:\u002F\u002Fgithub.com\u002Fjack-willturner\u002FDeepCompression-PyTorch) |\r\n\r\n## Related Repo\r\n\r\n[Awesome-model-compression-and-acceleration](https:\u002F\u002Fgithub.com\u002Fmemoiry\u002FAwesome-model-compression-and-acceleration)\r\n\r\n[EfficientDNNs](https:\u002F\u002Fgithub.com\u002FMingSun-Tse\u002FEfficientDNNs)\r\n\r\n[Embedded-Neural-Network](https:\u002F\u002Fgithub.com\u002FZhishengWang\u002FEmbedded-Neural-Network)\r\n\r\n[awesome-AutoML-and-Lightweight-Models](https:\u002F\u002Fgithub.com\u002Fguan-yuan\u002Fawesome-AutoML-and-Lightweight-Models)\r\n\r\n[Model-Compression-Papers](https:\u002F\u002Fgithub.com\u002Fchester256\u002FModel-Compression-Papers)\r\n\r\n[knowledge-distillation-papers](https:\u002F\u002Fgithub.com\u002Flhyfst\u002Fknowledge-distillation-papers)\r\n\r\n[Network-Speed-and-Compression](https:\u002F\u002Fgithub.com\u002Fmrgloom\u002FNetwork-Speed-and-Compression)\r\n","# 令人惊叹的剪枝 [![Awesome](https:\u002F\u002Fawesome.re\u002Fbadge.svg)](https:\u002F\u002Fawesome.re)\n\n一份精心整理的神经网络剪枝及相关资源列表。受 [awesome-deep-vision](https:\u002F\u002Fgithub.com\u002Fkjw0612\u002Fawesome-deep-vision)、[awesome-adversarial-machine-learning](https:\u002F\u002Fgithub.com\u002Fyenchenlin\u002Fawesome-adversarial-machine-learning)、[awesome-deep-learning-papers](https:\u002F\u002Fgithub.com\u002Fterryum\u002Fawesome-deep-learning-papers) 和 [Awesome-NAS](https:\u002F\u002Fgithub.com\u002FD-X-Y\u002FAwesome-NAS) 的启发而创建。\n\n欢迎随时提交 [pull request](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fawesome-Pruning\u002Fpulls) 或 [issue](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fawesome-Pruning\u002Fissues)，以添加新的论文。\n\n## 目录\n\n- [剪枝类型](#type-of-pruning)\n\n- [结构化剪枝综述](#a-survey-of-structured-pruning-arxiv-version-and-ieee-t-pami-version)\n\n- [2023年会议](#2023)\n\n- [2022年会议](#2022)\n\n- [2021年会议](#2021)\n\n- [2020年会议](#2020)\n\n- [2019年会议](#2019)\n\n- [2018年会议](#2018)\n\n- [2017年会议](#2017)\n\n- [2016年会议](#2016)\n\n- [2015年会议](#2015)\n\n### 剪枝类型\n\n| 类型 | `F` | `W` | `S` | `其他` |\n|:----------- |:--------------:|:--------------:|:----------------:|:-----------:|\n| 解释 | 滤波器剪枝 | 权重剪枝 | 特殊网络 | 其他类型 |\n\n### 结构化剪枝综述（[arXiv版本](https:\u002F\u002Farxiv.org\u002Fabs\u002F2303.00566) 和 [IEEE T-PAMI版本](https:\u002F\u002Fieeexplore.ieee.org\u002Fdocument\u002F10330640)）\n\n如果本文对您有所帮助，请引用我们的论文：\n```bibtex\n@article{he2024structured,\n author={He, Yang and Xiao, Lingao},\n journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, \n title={Deep卷积神经网络的结构化剪枝：综述}, \n year={2024},\n volume={46},\n number={5},\n pages={2900-2919},\n doi={10.1109\u002FTPAMI.2023.3334614}}\n```\n\n相关论文分类如下：\n![结构化剪枝分类法](https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhe-y_Awesome-Pruning_readme_834105cfd147.png)\n\n### 2023\n| 标题 | 会议地点 | 类型 | 编号 |\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\n| [通过拉马努金图的视角重新审视初始化时的剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=uVcDssQff_) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002Framanujan-on-pai)(即将发布) |\n| [揭秘彩票假设：获胜彩票掩码中编码了什么？](https:\u002F\u002Fopenreview.net\u002Fforum?id=xSsW2Am-ukZ) | ICLR | `W` | - |\n| [位剪枝：一种无需乘法的稀疏点积](https:\u002F\u002Fopenreview.net\u002Fforum?id=YUDiZcZTI8) | ICLR | `W` | [代码已删除](https:\u002F\u002Fgithub.com\u002FDensoITLab\u002Fbitprune) |\n| [NTK-SAP：通过对齐训练动态改进神经网络剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=-5EWhW_4qWP) | ICLR | `W` | - |\n| [软阈值剪枝的统一框架](https:\u002F\u002Fopenreview.net\u002Fforum?id=cCFqcrq0d8) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FYanqi-Chen\u002FLATS) |\n| [CrAM：一种考虑压缩的最小化器](https:\u002F\u002Fopenreview.net\u002Fforum?id=_eTZBs-yedr) | ICLR | `W` | - |\n| [保持可训练性的神经网络剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=AZFvpnnewr) | ICLR | `F` | - |\n| [DFPC：无需数据的耦合通道数据流驱动剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=mhnHqRqcjYU) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fdrive.google.com\u002Fdrive\u002Ffolders\u002F18eRYzWnB_6Qq0cYiSzvyOgicqn50g3-m) |\n| [TVSPrune——通过中间表示的总变差可分离性剪除非判别性滤波器，无需微调](https:\u002F\u002Fopenreview.net\u002Fforum?id=sZI1Oj9KBKy) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ftvsprune\u002FTVS_Prune) |\n| [HomoDistil：预训练Transformer的同伦任务无关蒸馏](https:\u002F\u002Fopenreview.net\u002Fforum?id=D7srTrGhAs) | ICLR | `F` | - |\n| [MECTA：内存经济型持续测试时模型适应](https:\u002F\u002Fopenreview.net\u002Fforum?id=N92hjSf5NNh) | ICLR | `F` | - |\n| [DepthFL：面向异构客户端的深度可分离联邦学习](https:\u002F\u002Fopenreview.net\u002Fforum?id=pf8RIZTMU58) | ICLR | `F` | - |\n| [OTOv2：自动、通用、用户友好](https:\u002F\u002Fopenreview.net\u002Fforum?id=7ynoX1ojPMt) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ftianyic\u002Fonly_train_once) |\n| [基于Polyak-Lojasiewicz条件的过参数化模型优化](https:\u002F\u002Fopenreview.net\u002Fforum?id=aBIpZvMdS56) | ICLR | `F` | - |\n| [从稀疏性视角看深度神经网络剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=i-DleYh34BM) | ICLR | `WF` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fdem123456789\u002FPruning-Deep-Neural-Networks-from-a-Sparsity-Perspective) |\n| [整体式对抗鲁棒剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=sAJDi9lD06L) | ICLR | `WF` | - |\n| [我如何学会不再担心并爱上再训练](https:\u002F\u002Fopenreview.net\u002Fforum?id=_nF5imFKQI) | ICLR | `WF` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FZIB-IOL\u002FBIMP) |\n| [量子神经网络中的对称剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=K96AogLDT2K) | ICLR | `S` | - |\n| [重新思考图上的彩票剪枝：图的稀疏性至关重要](https:\u002F\u002Fopenreview.net\u002Fforum?id=fjh7UGQgOB) | ICLR | `S` | - |\n| [边-模型联合稀疏学习对图神经网络而言在理论上是高效的](https:\u002F\u002Fopenreview.net\u002Fforum?id=4UldFtZ_CVF) | ICLR | `S` | - |\n| [图神经网络中的彩票剪枝搜索：双重视角](https:\u002F\u002Fopenreview.net\u002Fforum?id=Dvs-a3aymPe) | ICLR | `S` | - |\n| [基于拓扑排序的扩散模型用于因果发现](https:\u002F\u002Fopenreview.net\u002Fforum?id=Idusfje4-Wq) | ICLR | `S` | - |\n| [证明等变强彩票假设的一般框架](https:\u002F\u002Fopenreview.net\u002Fforum?id=vVJZtlZB9D) | ICLR | `其他` | - |\n| [稀疏性也许会哭泣：让我们一起让当前的稀疏神经网络失败吧！](https:\u002F\u002Fopenreview.net\u002Fforum?id=J6F3lLg4Kdp) | ICLR | `其他` | - |\n| [针对神经梯度的N:M稀疏性下的最小方差无偏估计](https:\u002F\u002Fopenreview.net\u002Fforum?id=vuD2xEtxZcj) | ICLR | `其他` | - |\n\n### 2022\n| 标题 | 会议\u002F期刊 | 类型 | 代码 |\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\n| [参数高效的掩码网络](https:\u002F\u002Fopenreview.net\u002Fforum?id=7rcuQ_V2GFg) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyueb17\u002FPEMN) |\n| [\"无损\"压缩深度神经网络：一种高维神经切空间核方法](https:\u002F\u002Fopenreview.net\u002Fforum?id=NaW6T93F34m) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FModel-Compression\u002FLossless_Compression) |\n| [损失可以是福音：将自监督语音表示路由到高效多语言和多任务语音处理](https:\u002F\u002Fopenreview.net\u002Fforum?id=2EUJ4e6H4OX) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FGATECH-EIC\u002FS3-Router) |\n| [脱节的模型：从傅里叶视角看分布偏移鲁棒性](https:\u002F\u002Fopenreview.net\u002Fforum?id=YZ-N-sejjwO) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fsarafridov\u002FRobustNets) |\n| [通过剪枝随机初始化网络获得稳健的二值化模型](https:\u002F\u002Fopenreview.net\u002Fforum?id=5g-h_DILemH) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FIVRL\u002FRobustBinarySubNet) |\n| [稀有彩票：在初始化时找到彩票剪枝方案](https:\u002F\u002Fopenreview.net\u002Fforum?id=Jpxd93u2vK-) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fksreenivasan\u002Fpruning_is_enough) |\n| [最优大脑压缩：一种精确的训练后量化与剪枝框架](https:\u002F\u002Fopenreview.net\u002Fforum?id=ksVGCOlOEba) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FIST-DASLab\u002FOBC) |\n| [通过训练与正则化的视角分析剪枝对泛化能力的影响](https:\u002F\u002Fopenreview.net\u002Fforum?id=OrcLKV9sKWp) | NeurIPS | `W` | - |\n| [Back Razor：通过自稀疏化反向传播实现内存高效的迁移学习](https:\u002F\u002Fopenreview.net\u002Fforum?id=mTXQIpXPDbh) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FBackRazor_Neurips22) |\n| [从PAC-贝叶斯理论视角分析彩票剪枝假设](https:\u002F\u002Fopenreview.net\u002Fforum?id=fbUybomIuE) | NeurIPS | `W` | - |\n| [稀疏的获胜彩票是数据高效的图像识别器](https:\u002F\u002Fopenreview.net\u002Fforum?id=wfKbtSjHA6F) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FDataEfficientLTH) |\n| [数据节食下的彩票剪枝：寻找具有稀疏可训练网络的初始化方案](https:\u002F\u002Fopenreview.net\u002Fforum?id=QLPzCpu756J) | NeurIPS | `W` | - |\n| [基于多样性驱动的模型压缩的加权互学习](https:\u002F\u002Fopenreview.net\u002Fforum?id=UQJoGBNRX4) | NeurIPS | `F` | - |\n| [SInGE：通过集成梯度估计神经元重要性实现稀疏化](https:\u002F\u002Fopenreview.net\u002Fforum?id=oQIJsMlyaW_) | NeurIPS | `F` | - |\n| [基于次模优化的数据高效结构化剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=K2QGzyLwpYG) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fmarwash25\u002Fsubpruning) |\n| [基于延迟-显著性背包问题的结构化剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=cUOR-_VsavA) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FHALP) |\n| [神经网络剪枝中的召回失真及未衰减剪枝算法](https:\u002F\u002Fopenreview.net\u002Fforum?id=5hgYi4r5MDp) | NeurIPS | `WF` | - |\n| [通过核心集与凸几何剪枝神经网络：迈向无假设的方法](https:\u002F\u002Fopenreview.net\u002Fforum?id=btpIaJiRx6z) | NeurIPS | `WF` | - |\n| [通过约束优化实现可控稀疏化：我如何学会停止调整惩罚项而爱上约束条件](https:\u002F\u002Fopenreview.net\u002Fforum?id=XUvSYc6TqDF) | NeurIPS | `WF` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fgallego-posada\u002Fconstrained_sparsity) |\n| [通过双层优化推进模型剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=t6O08FxvtBY) | NeurIPS | `WF` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FOPTML-Group\u002FBiP) |\n| [单片自组织脉冲神经元中层次化层的涌现](https:\u002F\u002Fopenreview.net\u002Fforum?id=cPVuuk1lZb3) | NeurIPS | `S` | - |\n| [CryptoGCN：快速且可扩展的同态加密图卷积网络推理](https:\u002F\u002Fopenreview.net\u002Fforum?id=VeQBBm1MmTZ) | NeurIPS | `S` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Franran0523\u002FCryptoGCN)(即将发布) |\n| [一次变换：频域中的高效算子学习](https:\u002F\u002Fopenreview.net\u002Fforum?id=B2PpZyAAEgV) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FDiffEqML\u002Fkairos)(即将发布) |\n| [大多数激活函数无需过深即可赢得彩票剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=NySDKS9SxN) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FRelationalML\u002FLT-existence) |\n| [剪枝对模型精度存在差异性影响](https:\u002F\u002Fopenreview.net\u002Fforum?id=11nMVZK0WYM) | NeurIPS | `其他` | - |\n| [神经网络的保模型压缩](https:\u002F\u002Fopenreview.net\u002Fforum?id=gt-l9Hu2ndd) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fjerry-chee\u002FModelPreserveCompressionNN) |\n| [蒸馏前先剪枝你的模型](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_8) | ECCV | `W` | [PyTorch(作者)](https:\u002F\u002Fhttps:\u002F\u002Fgithub.com\u002Fososos888\u002Fprune-then-distill) |\n| [FedLTN：用于稀疏且个性化的彩票剪枝网络的联邦学习](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19775-8_5) | ECCV | `W` | - |\n| [FairGRAPE：面向人脸属性分类的公平感知梯度剪枝方法](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19778-9_24) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FBernardo1998\u002FFairGRAPE) |\n| [超级彩票：通过联合架构搜索与参数剪枝从超网络中抽取任务无关的彩票剪枝方案](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_40) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FGATECH-EIC\u002FSuperTickets) |\n| [基于集成知识指导的子网络搜索与微调用于滤波器剪枝](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_34) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fsseung0703\u002FEKG) |\n| [编译器感知的模型剪枝：用于高效目标导向DNN执行](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20044-1_37) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ftaehokim20\u002FCPrune) |\n| [面向成本约束的通道剪枝软掩码](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_38) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FSMCP) |\n| [基于特征区分的深度神经网络滤波器剪枝](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19803-8_15) | ECCV | `F` | - |\n| [解耦的可微分网络剪枝](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_20) | ECCV | `F` | - |\n| [通过摊销推断的显著性图引导解释性网络剪枝](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19803-8_17) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FAlii-Ganjj\u002FInterpretationsSteeredPruning) |\n| [结合聚类与回滚的贝叶斯优化用于CNN自动剪枝](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20050-2_29) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ffanhanwei\u002FBOCR) |\n| [面向移动设备加速的多粒度剪枝](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_29) | ECCV | `WF` | - |\n| [在脉冲神经网络中探索彩票剪枝假设](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-19775-8_7) | ECCV | `S` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FIntelligent-Computing-Lab-Yale\u002FExploring-Lottery-Ticket-Hypothesis-in-SNNs) |\n| [利用时间剪枝迈向超低延迟的视觉与序列任务脉冲神经网络](https:\u002F\u002Flink.springer.com\u002F10.1007\u002F978-3-031-20083-0_42) | ECCV | `S` | - |\n| [神经网络初始化剪枝的最新进展](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F786) | IJCAI | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fmingsun-tse\u002Fsmile-pruning) |\n| [FedDUAP：使用服务器共享数据进行动态更新与适应性剪枝的联邦学习](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F385) | IJCAI | `F` | - |\n| [利用图卷积网络进行卷积神经网络加速的通道剪枝探讨](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F431) | IJCAI | `F` | - |\n| [剪枝即搜索：通过通道剪枝与结构重参数化实现高效的神经架构搜索](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F449) | IJCAI | `F` | - |\n| [通过合作协同进化进行神经网络剪枝](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2022\u002F667) | IJCAI | `F` | - |\n| [SPDY：兼具速度提升保证的精准剪枝](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Ffrantar22a.html) | ICML | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FIST-DASLab\u002Fspdy) |\n| [稀疏双重下降：网络剪枝如何加剧过拟合](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fhe22d.html) | ICML | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fhezheug\u002Fsparse-double-descent) |\n| [组合式脑外科医生：剪除神经网络中相互抵消的权重](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fyu22f.html) | ICML | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyuxwind\u002FCBS) |\n| [线性嫁接：放松的神经元剪枝有助于可认证的鲁棒性](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fchen22af.html) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FLinearity-Grafting) |\n| [提前赢得彩票：高效的早期网络剪枝](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Frachwan22a.html) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fjohnrachwan123\u002FEarly-Cropression-via-Gradient-Flow-Preservation) |\n| [基于多阶段图嵌入与强化学习的拓扑感知网络剪枝](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fyu22e.html) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyusx-swapp\u002FGNN-RL-Model-Compression) |\n| [仅使用整数离散流实现快速无损神经压缩](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fwang22a.html) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fthu-ml\u002FIODF) |\n| [DepthShrinker：一种新的压缩范式，旨在提升紧凑型神经网络的实际硬件效率](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Ffu22c.html) | ICML | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002FDepthShrinker) |\n| [PAC-Net：一种用于归纳迁移学习的模型剪枝方法](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fmyung22a.html) | ICML | `其他` | - |\n| [神经网络剪枝能去噪特征并使局部连接在视觉任务中显现](https:\u002F\u002Fproceedings.mlr.press\u002Fv162\u002Fpellegrini22a.html) | ICML | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fphiandark\u002FSiftingFeatures) |\n| [间隔剪枝：利用适应性滤波器表示改进稀疏CNN的训练](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FWimmer_Interspace_Pruning_Using_Adaptive_Filter_Representations_To_Improve_Training_of_CVPR_2022_paper.html) | CVPR | `W` | - |\n| [屏蔽对抗性破坏：寻找对抗性显著性以构建鲁棒且稀疏的网络](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FLee_Masking_Adversarial_Damage_Finding_Adversarial_Saliency_for_Robust_and_Sparse_CVPR_2022_paper.html) | CVPR | `W` | - |\n| [何时剪枝？一种面向早期结构化剪枝的策略](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FShen_When_To_Prune_A_Policy_Towards_Early_Structural_Pruning_CVPR_2022_paper.html) | CVPR | `F` | - |\n| [火在一起，线在一起：一种带有自监督掩码预测的动态剪枝方法](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FElkerdawy_Fire_Together_Wire_Together_A_Dynamic_Pruning_Approach_With_Self-Supervised_CVPR_2022_paper.html) | CVPR | `F` | - |\n| [重新审视随机通道剪枝用于神经网络压缩](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FLi_Revisiting_Random_Channel_Pruning_for_Neural_Network_Compression_CVPR_2022_paper.html) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fofsoundof\u002Frandom_channel_pruning)(即将发布) |\n| [利用完整经验回放学习贝叶斯稀疏网络以实现持续学习](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FYan_Learning_Bayesian_Sparse_Networks_With_Full_Experience_Replay_for_Continual_CVPR_2022_paper.html) | CVPR | `F` | - |\n| [DECORE：基于强化学习的深度压缩](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FAlwani_DECORE_Deep_Compression_With_Reinforcement_Learning_CVPR_2022_paper.html) | CVPR | `F` | - |\n| [CHEX：用于CNN模型压缩的通道探索](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FHou_CHEX_CHannel_EXploration_for_CNN_Model_Compression_CVPR_2022_paper.html) | CVPR | `F` | - |\n| [用少量样本压缩模型：模仿然后替换](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FWang_Compressing_Models_With_Few_Samples_Mimicking_Then_Replacing_CVPR_2022_paper.html) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fcjnjuwhy\u002FMiR)(即将发布) |\n| [对比双门控：利用对比学习学习稀疏特征](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FMeng_Contrastive_Dual_Gating_Learning_Sparse_Features_With_Contrastive_Learning_CVPR_2022_paper.html) | CVPR | `WF` | - |\n| [DiSparse：面向多任务模型压缩的解耦稀疏化](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2022\u002Fhtml\u002FSun_DiSparse_Disentangled_Sparsification_for_Multitask_Model_Compression_CVPR_2022_paper.html) | CVPR | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FSHI-Labs\u002FDiSparse-Multitask-Model-Compression) |\n| [通过Frank-Wolfe学习剪枝友好的网络：一次性、任意稀疏度且无需再训练](https:\u002F\u002Fopenreview.net\u002Fforum?id=O1DEtITim__) | ICLR **(Spotlight)** | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FSFW-Once-for-All-Pruning) |\n| [关于深度强化学习中的彩票剪枝与最小任务表示](https:\u002F\u002Fopenreview.net\u002Fforum?id=Fl3Mg_MZR-) | ICLR **(Spotlight)** | `W` | - |\n| [从算子理论视角看深度神经网络剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=pWBNOgdeURp) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fwilliam-redman\u002FKoopman_pruning) |\n| [通过计划性生长-剪枝方法实现有效的模型稀疏化](https:\u002F\u002Fopenreview.net\u002Fforum?id=xa6otUDdP2W) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fboone891214\u002FGaP) |\n| [签署超级掩码：保留、隐藏、反转](https:\u002F\u002Fopenreview.net\u002Fforum?id=e0jtGTfPihs) | ICLR | `W` | - |\n| [我们需要多少自由度来训练深度网络：从损失景观的角度来看](https:\u002F\u002Fopenreview.net\u002Fforum?id=ChMLTGRjFcU) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fganguli-lab\u002Fdegrees-of-freedom) |\n| [双重彩票剪枝假设](https:\u002F\u002Fopenreview.net\u002Fforum?id=fOsN52jn25l) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyueb17\u002FDLTH) |\n| [ peek-a-boo：随机加权神经网络中还隐藏着什么，以及如何高效地找到它](https:\u002F\u002Fopenreview.net\u002Fforum?id=moHCzz6D5H3) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FPeek-a-Boo) |\n| [稀疏两次胜出：更高效的训练带来更好的鲁棒泛化](https:\u002F\u002Fopenreview.net\u002Fforum?id=SYuJXrXq8tw) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FSparsity-Win-Robust-Generalization) |\n| [SOSP：通过二阶结构化剪枝高效捕捉全局相关性](https:\u002F\u002Fopenreview.net\u002Fforum?id=t5EmXZ3ZLR) | ICLR **(Spotlight)** | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fboschresearch\u002Fsosp)(即将发布) |\n| [像素化蝴蝶：简单高效的神经网络模型稀疏训练](https:\u002F\u002Fopenreview.net\u002Fforum?id=Nfl-iXa-y7R) | ICLR **(Spotlight)** | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FHazyResearch\u002Fpixelfly) |\n| [用彩票调控的组卷积重新审视内核剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=LdEhiMG9WLO) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FchoH\u002Flottery_regulated_grouped_kernel_pruning) |\n| [种下并寻找：你能找到那张获胜的彩票吗？](https:\u002F\u002Fopenreview.net\u002Fforum?id=9n9c8sf0xm) | ICLR | `F` | [PyTorch(作者)](http:\u002F\u002Fwww.github.com\u002FRelationalML\u002FPlantNSeek) |\n| [证明卷积神经网络的彩票剪枝假设](https:\u002F\u002Fopenreview.net\u002Fforum?id=Vjki79-619-) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FArthurWalraven\u002Fcnnslth) |\n| [关于通用彩票剪枝的存在性](https:\u002F\u002Fopenreview.net\u002Fforum?id=SYB4WrJql1n) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FRelationalML\u002FUniversalLT) |\n| [通过流形识别与方差减少训练结构化神经网络](https:\u002F\u002Fopenreview.net\u002Fforum?id=mdUYT5QV0O) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fwww.github.com\u002Fzihsyuan1214\u002Frmda) |\n| [通过结构化正则化剪枝学习高效的图像超分辨率网络](https:\u002F\u002Fopenreview.net\u002Fforum?id=AjGC97Aofee) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FMingSun-Tse\u002FSRP) |\n| [前景剪枝：利用元梯度在初始化时寻找可训练权重](https:\u002F\u002Fopenreview.net\u002Fforum?id=AIgn9uwfcD1) | ICLR | `WF` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fmil-ad\u002Fprospr) |\n| [随机剪枝的不合理有效性：最朴素的稀疏训练基线回归](https:\u002F\u002Fopenreview.net\u002Fforum?id=VBZJ_3tz-t) | ICLR | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FRandom_Pruning) |\n| [剪枝并调优集成：用稀疏独立子网络实现低成本集成学习](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F20842) | AAAI | `W` | - |\n| [先验梯度掩码引导的剪枝感知微调](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F19888) | AAAI | `F` | - |\n| [通过广义克罗内克积分解压缩卷积神经网络](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F19958) | AAAI | `其他` | - |\n\n### 2021 年\n| 标题 | 会议\u002F期刊 | 类型 | 代码 |\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\n| [利用惯性流形理论验证彩票假设](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Ffdc42b6b0ee16a2f866281508ef56730-Abstract.html) | NeurIPS | `W` | - |\n| [弹性彩票假设](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fdfccdb8b1cc7e4dab6d33db0fef12b88-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FElasticLTH) |\n| [彩票票券的合理性检验：你的中奖券真的能赢大奖吗？](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F6a130f1dc6f0c829f874e92e5458dced-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fboone891214\u002Fsanity-check-LTH) |\n| [为什么彩票票券会获胜？稀疏神经网络样本复杂度的理论视角](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F15f99f2165aa8c86c9dface16fefd281-Abstract.html) | NeurIPS | `W` | - |\n| [你偷了我的中奖彩票票券！让彩票票券声明其所有权](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F23e582ad8087f2c03a5a31c125123f9a-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FNO-stealing-LTH) |\n| [通过迭代随机化剪枝随机初始化的神经网络](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F23e582ad8087f2c03a5a31c125123f9a-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fdchiji-ntt\u002Fiterand) |\n| [通过神经再生增强剪枝可塑性实现稀疏训练](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F5227b6aaf294f5f027273aebf16015f2-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FGraNet) |\n| [AC\u002FDC：深度神经网络的压缩与解压缩交替训练](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F48000647b315f6f00f913caa757a70b3-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FIST-DASLab\u002FACDC) |\n| [制胜一手：压缩深度网络可以提高分布外鲁棒性](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F0607f4c705595b911a4f3e7a127b44e0-Abstract.html) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FRobustBench\u002Frobustbench) |\n| [重新思考卷积神经网络的剪枝标准](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F87ae6fb631f7c8a627e8e28785d9992d-Abstract.html) | NeurIPS | `F` | - |\n| [只需训练一次：一次性神经网络训练与剪枝框架](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fa376033f78e144f494bfc743c0be3330-Abstract.html) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ftianyic\u002Fonlytrainonce) |\n| [CHIP：基于通道独立性的紧凑神经网络剪枝](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fce6babd060aa46c61a5777902cca78af-Abstract.html) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FEclipsess\u002FCHIP_NeurIPS2021) |\n| [RED：寻找冗余以实现无数据结构化压缩深度神经网络](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fae5e3ce40e0404a45ecacaaf05e5f735-Abstract.html) | NeurIPS | `F` | - |\n| [神经网络压缩：迈向确定最优的逐层分解](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F2adcfc3929e7c03fac3100d3ad51da26-Abstract.html) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Flucaslie\u002Ftorchprune) |\n| [稀疏流：剪枝连续深度模型](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fbf1b2f4b901c21a1d8645018ea9aeb05-Abstract.html) | NeurIPS | `WF` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Flucaslie\u002Ftorchprune) |\n| [通过ReLU稳定性扩大精确神经网络压缩规模](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Fe35d7a5768c4b85b4780384d55dc3620-Abstract.html) | NeurIPS | `S` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyuxwind\u002FExactCompression) |\n| [GAN压缩中的判别器：生成器-判别器协同压缩方案](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Feffc299a1addb07e7089f9b269c31f2f-Abstract.html) | NeurIPS | `S` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FSJLeo\u002FGCC) |\n| [SGD中的重尾分布与过参数化神经网络的可压缩性](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002Ff5c3dd7514bf620a1b85450d2ae374b1-Abstract.html) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fmbarsbey\u002Fsgd_comp_gen) |\n| [ResRep：通过解耦记忆与遗忘实现无损CNN剪枝](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FDing_ResRep_Lossless_CNN_Pruning_via_Decoupling_Remembering_and_Forgetting_ICCV_2021_paper.html) | ICCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FDingXiaoH\u002FResRep) |\n| [通过神经架构搜索和剪枝实现移动端实时超分辨率](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FZhan_Achieving_On-Mobile_Real-Time_Super-Resolution_With_Neural_Architecture_and_Pruning_Search_ICCV_2021_paper.html) | ICCV | `F` | - |\n| [GDP：通过可微极化的门控实现稳定的神经网络剪枝](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FGuo_GDP_Stabilized_Neural_Network_Pruning_via_Gates_With_Differentiable_Polarization_ICCV_2021_paper.html) | ICCV | `F` | - |\n| [用于神经网络剪枝的自动图编码器-解码器](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FYu_Auto_Graph_Encoder-Decoder_for_Neural_Network_Pruning_ICCV_2021_paper.html) | ICCV | `F` | - |\n| [模型压缩中的探索与估计](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F5227b6aaf294f5f027273aebf16015f2-Abstract.html) | ICCV | `F` | - |\n| [子比特神经网络：学习压缩并加速二值神经网络](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FICCV2021\u002Fhtml\u002FWang_Sub-Bit_Neural_Networks_Learning_To_Compress_and_Accelerate_Binary_Neural_ICCV_2021_paper.html) | ICCV | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyikaiw\u002FSNN) |\n| [关于跨尺度剪枝的可预测性](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.10621) | ICML | `W` | - |\n| [神经网络剪枝的概率方法](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.10065) | ICML | `F` | - |\n| [从三个维度加速CNN：一个综合剪枝框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.04879) | ICML | `F` | - |\n| [面向实际网络压缩的分组Fisher剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2108.00708) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fjshilong\u002FFisherPruning) |\n| [通过协作式压缩迈向紧凑CNN](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.11228) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fliuguoyou\u002FTowards-Compact-CNNs-via-Collaborative-Compression) |\n| [置换、量化与微调：高效压缩神经网络的方法](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.15703) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fuber-research\u002Fpermute-quantize-finetune) |\n| [NPAS：面向超越实时移动加速的编译器感知统一网络剪枝与架构搜索框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F2012.00596) | CVPR | `F` | - |\n| [通过性能最大化进行网络剪枝](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent\u002FCVPR2021\u002Fhtml\u002FGao_Network_Pruning_via_Performance_Maximization_CVPR_2021_paper.html) | CVPR | `F` | - |\n| [通过减少结构冗余进行卷积神经网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.03438) | CVPR | `F` | - |\n| [流形正则化的动态网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2103.05861) | CVPR | `F` | - |\n| [Joint-DetNAS：用NAS、剪枝和动态蒸馏升级你的检测器](https:\u002F\u002Farxiv.org\u002Fabs\u002F2105.12971) | CVPR | `FO` | - |\n| [内容感知的GAN压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F2104.02244) | CVPR | `S` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Flychenyoko\u002Fcontent-aware-gan-compression) |\n| [多奖彩票假设：通过剪枝随机加权网络寻找准确的二值神经网络](https:\u002F\u002Fopenreview.net\u002Fforum?id=U_mat0b9iv) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fchrundle\u002Fbiprop) |\n| [基于幅度的剪枝的层自适应稀疏性](https:\u002F\u002Fopenreview.net\u002Fforum?id=H6ATjJ0TKdf) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fjaeho-lee\u002Flayer-adaptive-sparsity) |\n| [在初始化时剪枝神经网络：为什么我们总是错失目标？](https:\u002F\u002Fopenreview.net\u002Fforum?id=Ig-VyQc-MLK) | ICLR | `W` | - |\n| [初始化时的稳健剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=vXj_ucZQ4hA) | ICLR | `W` | - |\n| [用于分析网络剪枝的梯度流框架](https:\u002F\u002Fopenreview.net\u002Fforum?id=rumv7QmLUue) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FEkdeepSLubana\u002Fflowandprune) |\n| [通过增长正则化进行神经剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=o966_Is_nPA) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FMingSun-Tse\u002FRegularization-Pruning) |\n| [ChipNet：基于海维赛德连续近似的预算感知剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=xCxXwTzx4L1) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FtransmuteAI\u002FChipNet) |\n| [有意义的网络剪枝：再训练变体案例研究](https:\u002F\u002Fopenreview.net\u002Fforum?id=Cb54AMqHQFP) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Flehduong\u002FNPTM) |\n\n### 2020年\r\n\r\n| 标题 | 会议\u002F期刊 | 类型 | 编号 |\r\n|:-------------------------------------------------------------------------------------------------------------------------------- |:-----:|:-------:|:----:|\r\n| [通过子集和寻找最优彩票通票：对数级过参数化就足够了](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2020\u002Fhash\u002F1b742ae215adf18b75449c6e272fd92d-Abstract.html) | NeurIPS | `W` | - |\r\n| [利用连续稀疏化赢得彩票大奖](https:\u002F\u002Farxiv.org\u002Fabs\u002F1912.04427v4) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Flolemacs\u002Fcontinuous-sparsification) |\r\n| [HYDRA：剪枝对抗鲁棒的神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.10509) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Finspire-group\u002Fhydra) |\r\n| [对数稀疏化就是你需要的一切](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.12156) | NeurIPS | `W` | - |\r\n| [深度神经网络的方向性剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.09358) | NeurIPS | `W` | - |\r\n| [运动剪枝：通过微调实现自适应稀疏化](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.07683) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fblock_movement_pruning) |\r\n| [剪枝方法的合理性检验：随机通票也能中大奖](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.11094) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FJingtongSu\u002Fsanity-checking-pruning) |\r\n| [神经元合并：弥补被剪掉的神经元](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.13160) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ffriendshipkim\u002Fneuron-merging) |\r\n| [基于极化正则化的神经元级结构化剪枝](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2020\u002Ffile\u002F703957b6dd9e3a7980e040bee50ded65-Paper.pdf) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fpolarizationpruning\u002FPolarizationPruning) |\r\n| [SCOP：用于可靠神经网络剪枝的科学控制方法](https:\u002F\u002Farxiv.org\u002Fabs\u002F2010.10732) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyehuitang\u002FPruning\u002Ftree\u002Fmaster\u002FSCOP_NeurIPS2020) |\r\n| [基于深度强化学习的存储高效、动态灵活的运行时通道剪枝](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2020\u002Fhash\u002Fa914ecef9c12ffdb9bede64bb703d877-Abstract.html) | NeurIPS | `F` | - |\r\n| [神经网络剪枝中的泛化与稳定性权衡](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.03728) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fbbartoldson\u002FGeneralizationStabilityTradeoff) |\r\n| [贪心优化可证明地赢得彩票大奖：对数数量的幸运通票就足够了](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2020\u002Fhash\u002Fbe23c41621390a448779ee72409e5f49-Abstract.html) | NeurIPS | `WF` | - |\r\n| [滤波器内的滤波器剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2009.14410) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Ffxmeng\u002FPruning-Filter-in-Filter) |\r\n| [基于位置的缩放梯度用于模型量化和剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11035) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FJangho-Kim\u002FPSG-pytorch) |\r\n| [贝叶斯比特：统一量化与剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.07093) | NeurIPS | `其他` | - |\r\n| [通过迭代保持突触流，在无需任何数据的情况下剪枝神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.05467) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fganguli-lab\u002FSynaptic-Flow) |\r\n| [结合网络剪枝的元学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.03219) | ECCV | `W` | - |\r\n| [通过剪枝激活梯度加速CNN训练](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.00173) | ECCV | `W` | - |\r\n| [EagleEye：用于高效神经网络剪枝的快速子网评估](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.02491) | ECCV **(口头报告)** | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fanonymous47823493\u002FEagleEye) |\r\n| [DSA：通过可微稀疏分配实现更高效的预算剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2004.02164) | ECCV | `F` | - |\r\n| [DHP：基于超网络的可微元剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.13683) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fofsoundof\u002Fdhp) |\r\n| [DA-NAS：面向高效神经架构搜索的数据自适应剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.12563) S | ECCV | `其他` | - |\r\n| [面向硬件效率的可微联合剪枝与量化](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.10463) | ECCV | `其他` | - |\r\n| [通过自动结构搜索进行通道剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2001.08565) | IJCAI | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Flmbxmu\u002FABCPruner) |\r\n| [具有潜在漏洞抑制功能的对抗性神经剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.04355) | ICML | `W` | - |\r\n| [证明彩票通票假设：剪枝就是你需要的一切](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.00585) | ICML | `W` | - |\r\n| [通过贪婪子网络选择进行网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.01794) | ICML | `F` | - |\r\n| [使用可微掩码进行操作感知的软通道剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2007.03938) | ICML | `F` | - |\r\n| [DropNet：通过迭代剪枝降低神经网络复杂度](https:\u002F\u002Fproceedings.mlr.press\u002Fv119\u002Ftan20a.html) | ICML | `F` | - |\r\n| [用于可学习稀疏性的软阈值权重重参数化](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.03231) | ICML | `WF` | [Pytorch(作者)](https:\u002F\u002Fgithub.com\u002FRAIVNLab\u002FSTR) |\r\n| [通过权重加密实现结构化压缩，用于非结构化剪枝和量化](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.10138) | CVPR | `W` | - |\r\n| [基于稀疏-量化联合学习的自动神经网络压缩：一种基于约束优化的方法](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fpapers\u002FYang_Automatic_Neural_Network_Compression_by_Sparsity-Quantization_Joint_Learning_A_Constrained_CVPR_2020_paper.pdf) | CVPR | `W` | - |\r\n| [通过学习全局排名实现高效模型压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.12368) | CVPR **(口头报告)** | `F` | [Pytorch(作者)](https:\u002F\u002Fgithub.com\u002Fcmu-enyac\u002FLeGR) |\r\n| [HRank：利用高秩特征图进行滤波器剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.10179) | CVPR **(口头报告)** | `F` | [Pytorch(作者)](https:\u002F\u002Fgithub.com\u002Flmbxmu\u002FHRank) |\r\n| [带有残差连接且数据有限的神经网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.08114) | CVPR **(口头报告)** | `F` | - |\r\n| [DMCP：面向神经网络的可微马尔可夫通道剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.03354) | CVPR **(口头报告)** | `F` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Fzx55\u002Fdmcp) |\r\n| [组稀疏性：滤波器剪枝与分解之间的桥梁，用于网络压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.08935) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fofsoundof\u002Fgroup_sparsity) |\r\n| [少量样本知识蒸馏用于高效网络压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.01839) | CVPR | `F` | - |\r\n| [面向深度神经网络的资源受限离散模型压缩](http:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fhtml\u002FGao_Discrete_Model_Compression_With_Resource_Constraint_for_Deep_Neural_Networks_CVPR_2020_paper.html) | CVPR | `F` | - |\r\n| [为加速深度卷积神经网络而学习滤波器剪枝标准](http:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fhtml\u002FHe_Learning_Filter_Pruning_Criteria_for_Deep_Convolutional_Neural_Networks_Acceleration_CVPR_2020_paper.html) | CVPR | `F` | - |\r\n| [APQ：联合搜索网络架构、剪枝策略和量化策略](https:\u002F\u002Farxiv.org\u002Fabs\u002F2006.08509) | CVPR | `F` | - |\r\n| [多维剪枝：模型压缩的统一框架](http:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2020\u002Fhtml\u002FGuo_Multi-Dimensional_Pruning_A_Unified_Framework_for_Model_Compression_CVPR_2020_paper.html) | CVPR **(口头报告)** | `WF` | - |\r\n| [从信号传播角度在初始化时剪枝神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.06307) | ICLR **(亮点论文)** | `W` | - |\r\n| [ProxSGD：在正则化和约束条件下训练结构化神经网络](https:\u002F\u002Fopenreview.net\u002Fforum?id=HygpthEtvr) | ICLR | `W` | [TF+PT(作者)](https:\u002F\u002Fgithub.com\u002Foptyang\u002Fproxsgd) |\r\n| [通过雅可比谱评估一次性剪枝循环神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1912.00120) | ICLR | `W` | - |\r\n| [前瞻：一种基于幅度的剪枝的远见替代方案](https:\u002F\u002Farxiv.org\u002Fabs\u002F2002.04809) | ICLR | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Falinlab\u002Flookahead_pruning) |\r\n| [通过核集进行数据无关的神经剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.04018) | ICLR | `W` | - |\r\n| [可证明的高效神经网络滤波器剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1911.07412) | ICLR | `F` | - |\r\n| [带有反馈的动态模型剪枝](https:\u002F\u002Fopenreview.net\u002Fforum?id=SJem8lSFwB) | ICLR | `WF` | - |\r\n| [比较神经网络剪枝中的回溯与微调](https:\u002F\u002Farxiv.org\u002Fabs\u002F2003.02389) | ICLR **(口头报告)** | `WF` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Flottery-ticket\u002Frewinding-iclr20-public) |\r\n| [AutoCompress：一个用于超高压缩率的自动DNN结构化剪枝框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.03141) | AAAI | `F` | - |\r\n| [重生的滤波器：用有限数据剪枝卷积神经网络](https:\u002F\u002Fojs.aaai.org\u002Findex.php\u002FAAAI\u002Farticle\u002Fview\u002F6058) | AAAI | `F` | - |\r\n| [DARB：一种面向深度神经网络的密度感知常规块剪枝](http:\u002F\u002Farxiv.org\u002Fabs\u002F1911.08020) | AAAI | `其他` | - |\r\n| [从零开始的剪枝](http:\u002F\u002Farxiv.org\u002Fabs\u002F1909.12579) | AAAI | `其他` | - |\n\n### 2019年\r\n\r\n| 标题 | 会议\u002F期刊 | 类型 | 编号 |\r\n|:-------|:--------:|:-------:|:-------:|\r\n| [拆解彩票彩票：零、符号与超级掩码](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.01067) | NeurIPS | `W` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Fuber-research\u002Fdeconstructing-lottery-tickets) |\r\n| [一张彩票通吃所有：跨数据集和优化器的彩票初始化泛化](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.02773) | NeurIPS | `W` | - |\r\n| [用于剪枝超深神经网络的全局稀疏动量SGD](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.12778) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FDingXiaoH\u002FGSM-SGD) |\r\n| [AutoPrune：通过正则化辅助参数实现自动网络剪枝](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F9521-autoprune-automatic-network-pruning-by-regularizing-auxiliary-parameters) | NeurIPS | `W` | - |\r\n| [基于可变换架构搜索的网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.09717) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FD-X-Y\u002FNAS-Projects) |\r\n| [Gate Decorator：加速深度卷积神经网络的全局滤波器剪枝方法](https:\u002F\u002Farxiv.org\u002Fabs\u002F1909.08174) | NeurIPS | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyouzhonghui\u002Fgate-decorator-pruning) |\r\n| [结合对抗鲁棒性的模型压缩：统一的优化框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F1902.03538) | NeurIPS | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FTAMU-VITA\u002FATMC) |\r\n| [对抗鲁棒性与模型压缩，二者兼得？](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.12561) | ICCV | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fyeshaokai\u002FRobustness-Aware-Pruning-ADMM) |\r\n| [MetaPruning：用于自动神经网络通道剪枝的元学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.10258) | ICCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fliuzechun\u002FMetaPruning) |\r\n| [通过递归贝叶斯剪枝加速CNN](https:\u002F\u002Farxiv.org\u002Fabs\u002F1812.00353) | ICCV | `F` | - |\r\n| [用于卷积神经网络压缩的滤波器基学习](https:\u002F\u002Farxiv.org\u002Fabs\u002F1908.08932) | ICCV | `其他` | - |\r\n| [用于无配对图像翻译的共进化压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F1907.10804) | ICCV | `S` | - |\r\n| [COP：基于相关性正则化的滤波器级剪枝实现定制化深度模型压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F1906.10337) | IJCAI | `F` | [Tensorflow(作者)](https:\u002F\u002Fgithub.com\u002FZJULearning\u002FCOP) |\r\n| [基于几何中位数的滤波器剪枝用于加速深度卷积神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.00250) | CVPR **(口头报告)** | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Ffilter-pruning-geometric-median) |\r\n| [通过生成对抗学习实现最优结构化CNN剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.09291) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FShaohuiLin\u002FGAL) |\r\n| [向心SGD用于剪枝结构复杂的超深卷积网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1904.03837) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FShawnDing1994\u002FCentripetal-SGD) |\r\n| [关于卷积神经网络中的隐式滤波器级稀疏性](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.12495)，[扩展1](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.04967)，[扩展2](https:\u002F\u002Fopenreview.net\u002Fforum?id=rylVvNS3hE) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fmehtadushy\u002FSelecSLS-Pytorch) |\r\n| [基于预算感知正则化的神经网络结构化剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1811.09332) | CVPR | `F` | - |\r\n| [用于神经网络剪枝的重要性估计](http:\u002F\u002Fjankautz.com\u002Fpublications\u002FImportance4NNPruning_CVPR19.pdf) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FNVlabs\u002FTaylor_pruning) |\r\n| [OICSR：用于紧凑型深度神经网络的外–内–通道稀疏性正则化](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.11664) | CVPR | `F` | - |\r\n| [变分卷积神经网络剪枝](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_CVPR_2019\u002Fhtml\u002FZhao_Variational_Convolutional_Neural_Network_Pruning_CVPR_2019_paper.html) | CVPR | `F` | - |\r\n| [偏序剪枝：在神经架构搜索中实现最佳速度\u002F精度权衡](https:\u002F\u002Farxiv.org\u002Fabs\u002F1903.03777) | CVPR | `其他` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Flixincn2015\u002FPartial-Order-Pruning) |\r\n| [深度网络的协作式通道剪枝](http:\u002F\u002Fproceedings.mlr.press\u002Fv97\u002Fpeng19c.html) | ICML | `F` | - |\r\n| [近似Oracle滤波器剪枝用于破坏性CNN宽度优化](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.04748) | ICML | `F` | - |\r\n| [EigenDamage：在克罗内克分解特征基下的结构化剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1905.05934) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Falecwangcq\u002FEigenDamage-Pytorch) |\r\n| [彩票假设：寻找稀疏且可训练的神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1803.03635) | ICLR **(最佳论文)** | `W` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Fgoogle-research\u002Flottery-ticket-hypothesis) |\r\n| [SNIP：基于连接敏感性的单次网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.02340) | ICLR | `W` | [TensorFLow(作者)](https:\u002F\u002Fgithub.com\u002Fnamhoonlee\u002Fsnip-public) |\r\n| [动态通道剪枝：特征增强与抑制](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.05331) | ICLR | `F` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Fdeep-fry\u002Fmayo) |\r\n| [重新思考网络剪枝的价值](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.05270) | ICLR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FEric-mingjie\u002Frethinking-network-pruning) |\r\n| [用于高效深度学习的动态稀疏图](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.00859) | ICLR | `F` | [CUDA(第三方)](https:\u002F\u002Fgithub.com\u002Fmtcrawshaw\u002Fdynamic-sparse-graph) |\n\n### 2018 年\n| 标题 | 会议\u002F期刊 | 类型 | 代码 |\n|:-------|:--------:|:-------:|:-------:|\n| [卷积神经网络的频域动态剪枝](https:\u002F\u002Fpapers.NeurIPS.cc\u002Fpaper\u002F7382-frequency-domain-dynamic-pruning-for-convolutional-neural-networks.pdf) | NeurIPS | `W` | - |\n| [面向深度神经网络的判别感知通道剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1810.11809) | NeurIPS | `F` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002FSCUT-AILab\u002FDCP) |\n| [通过敏感性驱动正则化学习稀疏神经网络](https:\u002F\u002Farxiv.org\u002Fpdf\u002F1810.11764.pdf) | NeurIPS | `WF` | - |\n| [约束感知的深度神经网络压缩](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_ECCV_2018\u002Fhtml\u002FChangan_Chen_Constraints_Matter_in_ECCV_2018_paper.html) | ECCV | `W` | [SkimCaffe(作者)](https:\u002F\u002Fgithub.com\u002FChanganVR\u002FConstraintAwareCompression) |\n| [基于交替方向乘子法的系统性 DNN 权重剪枝框架](https:\u002F\u002Farxiv.org\u002Fabs\u002F1804.03294) | ECCV | `W` | [Caffe(作者)](https:\u002F\u002Fgithub.com\u002FKaiqiZhang\u002Fadmm-pruning) |\n| [AMC：用于移动端模型压缩与加速的 AutoML](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.03494) | ECCV | `F` | [TensorFlow(第三方)](https:\u002F\u002Fgithub.com\u002FTencent\u002FPocketFlow#channel-pruning) |\n| [面向深度神经网络的数据驱动稀疏结构选择](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.01213) | ECCV | `F` | [MXNet(作者)](https:\u002F\u002Fgithub.com\u002FTuSimple\u002Fsparse-structure-selection) |\n| [基于核心集的神经网络压缩](https:\u002F\u002Farxiv.org\u002Fabs\u002F1807.09810) | ECCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fmetro-smiles\u002FCNN_Compression) |\n| [用于加速深度卷积神经网络的软过滤器剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1808.06866) | IJCAI | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fsoft-filter-pruning) |\n| [通过全局与动态过滤器剪枝加速卷积网络](https:\u002F\u002Fwww.ijcai.org\u002Fproceedings\u002F2018\u002F0336.pdf) | IJCAI | `F` | - |\n| [Weightless：用于深度神经网络压缩的有损权重编码](https:\u002F\u002Fproceedings.mlr.press\u002Fv80\u002Freagan18a.html) | ICML | `W` | - |\n| [利用变分信息瓶颈压缩神经网络](https:\u002F\u002Fproceedings.mlr.press\u002Fv80\u002Fdai18d.html) | ICML | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fzhuchen03\u002FVIBNet) |\n| [Deep k-Means：通过更严格的聚类分配进行再训练和参数共享以压缩深度卷积网络](https:\u002F\u002Fproceedings.mlr.press\u002Fv80\u002Fwu18h.html) | ICML | `其他` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FVITA-Group\u002FDeep-K-Means-pytorch) |\n| [CLIP-Q：通过并行剪枝-量化实现深度网络压缩学习](https:\u002F\u002Fopenaccess.thecvf.com\u002Fcontent_cvpr_2018\u002Fhtml\u002FTung_CLIP-Q_Deep_Network_CVPR_2018_paper.html) | CVPR | `W` | - |\n| [用于神经网络剪枝的“学习-压缩”算法](http:\u002F\u002Ffaculty.ucmerced.edu\u002Fmcarreira-perpinan\u002Fpapers\u002Fcvpr18.pdf) | CVPR | `W` | - |\n| [PackNet：通过迭代剪枝将多个任务添加到单个网络中](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.05769) | CVPR | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Farunmallya\u002Fpacknet) |\n| [NISP：使用神经元重要性评分传播进行网络剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1711.05908) | CVPR | `F` | - |\n| [剪枝还是不剪枝：探索剪枝在模型压缩中的有效性](https:\u002F\u002Farxiv.org\u002Fabs\u002F1710.01878) | ICLR | `W` | - |\n| [重新思考卷积层通道剪枝中小范数即低信息量的假设](https:\u002F\u002Farxiv.org\u002Fabs\u002F1802.00124) | ICLR | `F` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002Fbobye\u002Fbatchnorm_prune), [PyTorch(第三方)](https:\u002F\u002Fgithub.com\u002Fjack-willturner\u002Fbatchnorm-pruning) |\n\n### 2017年\n| 标题 | 会议 | 类型 | 代码 |\n|:-------|:--------:|:-------:|:-------:|\n| [Net-Trim：具有性能保证的深度神经网络凸剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05162) | NeurIPS | `W` | [TensorFlow(作者)](https:\u002F\u002Fgithub.com\u002FDNNToolBox\u002FNet-Trim-v1) |\n| [通过逐层最优脑外科手术学习剪枝深度神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1705.07565) | NeurIPS | `W` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002Fcsyhhu\u002FL-OBS) |\n| [运行时神经网络剪枝](https:\u002F\u002Fpapers.NeurIPS.cc\u002Fpaper\u002F6813-runtime-neural-pruning) | NeurIPS | `F` | - |\n| [基于对数正态乘性噪声的结构化贝叶斯剪枝](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2017\u002Fhash\u002Fdab49080d80c724aad5ebf158d63df41-Abstract.html) | NeurIPS | `F` | - |\n| [深度学习中的贝叶斯压缩](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2017\u002Fhash\u002F69d1fc78dbda242c43ad6590368912d4-Abstract.html) | NeurIPS | `F` | - |\n| [ThiNet：一种用于深度神经网络压缩的滤波器级剪枝方法](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.06342) | ICCV | `F` | [Caffe(作者)](https:\u002F\u002Fgithub.com\u002FRoll920\u002FThiNet), [PyTorch(第三方)](https:\u002F\u002Fgithub.com\u002Ftranorrepository\u002Freprod-thinet) |\n| [用于加速超深神经网络的通道剪枝](https:\u002F\u002Farxiv.org\u002Fabs\u002F1707.06168) | ICCV | `F` | [Caffe(作者)](https:\u002F\u002Fgithub.com\u002Fyihui-he\u002Fchannel-pruning) |\n| [通过网络瘦身学习高效的卷积网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1708.06519) | ICCV | `F` | [PyTorch(作者)](https:\u002F\u002Fgithub.com\u002FEric-mingjie\u002Fnetwork-slimming) |\n| [变分丢弃稀疏化深度神经网络](http:\u002F\u002Farxiv.org\u002Fabs\u002F1701.05369) | ICML | `W` | - |\n| [深度神经网络的组合分组与互斥稀疏化](https:\u002F\u002Fproceedings.mlr.press\u002Fv70\u002Fyoon17a.html) | ICML | `WF` | - |\n| [利用能耗感知剪枝设计节能卷积神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.05128) | CVPR | `W` | - |\n| [为高效卷积神经网络剪枝滤波器](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.08710) | ICLR | `F` | [PyTorch(第三方)](https:\u002F\u002Fgithub.com\u002FEric-mingjie\u002Frethinking-network-pruning\u002Ftree\u002Fmaster\u002Fimagenet\u002Fl1-norm-pruning) |\n| [为资源高效推理而剪枝卷积神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1611.06440) | ICLR | `F` | [TensorFlow(第三方)](https:\u002F\u002Fgithub.com\u002FTencent\u002FPocketFlow#channel-pruning) |\n\n\n### 2016年\n| 标题 | 会议 | 类型 | 代码 |\n|:-------|:--------:|:-------:|:-------:|\n| [用于高效DNN的动态网络手术](https:\u002F\u002Farxiv.org\u002Fabs\u002F1608.04493) | NeurIPS | `W` | [Caffe(作者)](https:\u002F\u002Fgithub.com\u002Fyiwenguo\u002FDynamic-Network-Surgery) |\n| [学习深度网络中的神经元数量](https:\u002F\u002Fproceedings.neurips.cc\u002Fpaper\u002F2016\u002Fhash\u002F6e7d2da6d3953058db75714ac400b584-Abstract.html) | NeurIPS | `F` | - |\n| [深度压缩：通过剪枝、训练量化和霍夫曼编码压缩深度神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1510.00149) | ICLR **(最佳)** | `W` | [Caffe(作者)](https:\u002F\u002Fgithub.com\u002Fsonghan\u002FDeep-Compression-AlexNet) |\n\n\n### 2015年\n\n| 标题 | 会议 | 类型 | 代码 |\n|:-------|:--------:|:-------:|:-------:|\n| [同时学习权重和连接以构建高效神经网络](https:\u002F\u002Farxiv.org\u002Fabs\u002F1506.02626) | NeurIPS | `W` | [PyTorch(第三方)](https:\u002F\u002Fgithub.com\u002Fjack-willturner\u002FDeepCompression-PyTorch) | \n\n## 相关仓库\n\n[Awesome-model-compression-and-acceleration](https:\u002F\u002Fgithub.com\u002Fmemoiry\u002FAwesome-model-compression-and-acceleration)\n\n[EfficientDNNs](https:\u002F\u002Fgithub.com\u002FMingSun-Tse\u002FEfficientDNNs)\n\n[Embedded-Neural-Network](https:\u002F\u002Fgithub.com\u002FZhishengWang\u002FEmbedded-Neural-Network)\n\n[awesome-AutoML-and-Lightweight-Models](https:\u002F\u002Fgithub.com\u002Fguan-yuan\u002Fawesome-AutoML-and-Lightweight-Models)\n\n[Model-Compression-Papers](https:\u002F\u002Fgithub.com\u002Fchester256\u002FModel-Compression-Papers)\n\n[knowledge-distillation-papers](https:\u002F\u002Fgithub.com\u002Flhyfst\u002Fknowledge-distillation-papers)\n\n[Network-Speed-and-Compression](https:\u002F\u002Fgithub.com\u002Fmrgloom\u002FNetwork-Speed-and-Compression)","# Awesome-Pruning 快速上手指南\n\n**Awesome-Pruning** 并非一个单一的可安装软件包或框架，而是一个**精选的神经网络剪枝（Neural Network Pruning）资源列表**。它汇集了该领域的经典论文、综述文章以及对应的开源代码实现。\n\n本指南将指导开发者如何利用该列表快速定位所需算法，并运行相关的开源代码。\n\n## 1. 环境准备\n\n由于列表中包含多种不同年份和类型的剪枝算法（如滤波器剪枝 `F`、权重剪枝 `W`、特殊网络剪枝 `S` 等），具体的系统要求取决于你选择的特定项目。以下是通用的推荐配置：\n\n* **操作系统**: Linux (Ubuntu 18.04\u002F20.04+) 或 macOS\n* **编程语言**: Python 3.7+\n* **核心依赖**:\n * PyTorch (大多数现代剪枝算法基于 PyTorch)\n * TensorFlow (部分早期项目可能使用)\n* **硬件建议**: NVIDIA GPU (CUDA 支持) 用于加速模型训练和剪枝过程\n\n**前置知识**:\n建议先阅读列表中的综述文章以建立理论框架：\n* **结构化剪枝综述**: [Structured Pruning for Deep Convolutional Neural Networks: A Survey](https:\u002F\u002Farxiv.org\u002Fabs\u002F2303.00566) (IEEE T-PAMI 2024)\n\n## 2. 获取资源与安装\n\n由于 Awesome-Pruning 是资源索引，你需要先克隆仓库获取列表，然后根据需求跳转到具体项目的仓库进行安装。\n\n### 步骤 1: 克隆资源列表\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fhe-y\u002Fawesome-Pruning.git\ncd awesome-Pruning\n```\n\n### 步骤 2: 选择并安装具体算法\n在 `README.md` 中找到你感兴趣的论文（例如 2023 年 ICLR 的 `OTOv2` 或 `TVSPrune`），点击 \"Code\" 列中的链接进入原作者的代码仓库。\n\n**通用安装示例（以 PyTorch 项目为例）：**\n\n1. **创建虚拟环境** (推荐):\n ```bash\n python -m venv pruning_env\n source pruning_env\u002Fbin\u002Factivate # Windows 使用: pruning_env\\Scripts\\activate\n ```\n\n2. **安装 PyTorch** (推荐使用国内镜像源加速):\n ```bash\n # 使用清华大学镜像源安装 PyTorch (根据实际 CUDA 版本调整)\n pip install torch torchvision torchaudio --index-url https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n ```\n\n3. **安装项目依赖**:\n 进入你选定的具体项目目录后，通常执行：\n ```bash\n pip install -r requirements.txt\n ```\n *注：如果该项目未提供 `requirements.txt`，请参考其 README 中的具体安装指令。*\n\n## 3. 基本使用\n\n使用流程通常分为三个阶段：**复现论文环境** -> **加载预训练模型** -> **执行剪枝**。\n\n以下以典型的基于 PyTorch 的剪枝项目为例（具体命令需参照所选项目的官方文档）：\n\n### 场景：对 ResNet-50 进行滤波器剪枝 (Filter Pruning)\n\n1. **准备数据与模型**:\n 确保已下载 ImageNet 或其他目标数据集，并准备好标准的 ResNet-50 预训练权重。\n\n2. **运行剪枝脚本**:\n 大多数项目会提供类似的训练\u002F剪枝脚本。以下是一个概念性示例：\n\n ```bash\n # 示例命令：启动剪枝训练\n # --arch: 网络架构\n # --data: 数据集路径\n # --prune_ratio: 剪枝比例\n python main.py \\\n --arch resnet50 \\\n --data \u002Fpath\u002Fto\u002Fimagenet \\\n --prune_ratio 0.5 \\\n --save_dir .\u002Fpruned_model\n ```\n\n3. **评估剪枝后模型**:\n 剪枝完成后，通常需要对模型进行微调（Fine-tuning）以恢复精度，然后进行测试：\n\n ```bash\n python eval.py \\\n --resume .\u002Fpruned_model\u002Fcheckpoint.pth.tar \\\n --data \u002Fpath\u002Fto\u002Fimagenet\n ```\n\n### 关键分类参考\n在使用列表时，请根据任务需求参考以下类型标识：\n* **`F` (Filter pruning)**: 滤波器\u002F通道剪枝，直接减少计算量 (FLOPs)，适合部署加速。\n* **`W` (Weight pruning)**: 权重剪枝，主要减少参数量，通常需要专用稀疏计算库才能加速。\n* **`S` (Special Networks)**: 针对图神经网络 (GNN)、量子神经网络等特殊架构的剪枝。\n* **`Other`**: 其他相关技术，如彩票假设 (Lottery Ticket Hypothesis) 分析等。\n\n> **提示**: 列表中每个条目都附带了论文链接和代码链接。对于中国开发者，若原代码托管在 Google Drive 且下载困难，可尝试在 GitHub Issues 中搜索是否有国内开发者提供的备份，或使用学术加速工具下载论文。","某边缘计算团队正致力于将大型目标检测模型部署到算力受限的无人机上，急需在保持精度的同时大幅压缩模型体积。\n\n### 没有 Awesome-Pruning 时\n- **文献检索如大海捞针**：团队成员需在 Google Scholar 和 arXiv 上手动搜索\"pruning\"、\"compression\"等关键词，面对成千上万篇论文难以辨别哪些是真正有效的结构化剪枝方案。\n- **技术路线选择盲目**：由于缺乏系统的分类指引，团队难以区分“滤波器剪枝（Filter pruning）”与“权重剪枝（Weight pruning）”的具体适用场景，导致初期选型错误，浪费数周时间复现不兼容硬件的算法。\n- **代码资源分散难寻**：即使找到了理论可行的论文，往往需要花费大量精力去各个作者的个人页或隐藏深的仓库中寻找开源代码，甚至常遇到代码缺失或无法运行的情况。\n- **前沿动态跟进滞后**：团队仅关注几年前的经典方法，错过了 2023 年 ICLR 等顶会上关于初始化剪枝或动态对齐的最新突破，导致最终模型压缩率未达预期。\n\n### 使用 Awesome-Pruning 后\n- **一站式获取权威资源**：直接查阅按年份和会议（如 ICLR、CVPR）整理的清单，迅速锁定近年最高质量的剪枝论文，将文献调研时间从数周缩短至几天。\n- **清晰的技术分类导航**：利用其详细的“剪枝类型”表格，团队快速明确了适合无人机硬件加速的结构化剪枝路线，避免了在非结构化稀疏方案上的无效尝试。\n- **代码链接直达可用**：通过列表中提供的官方代码仓库链接（如 PyTorch 实现），团队直接复现了\"Revisiting Pruning at Initialization\"等前沿算法，显著提升了开发效率。\n- **紧跟领域最新进展**：借助持续更新的 2023-2024 年新论文列表，团队及时引入了基于训练动态对齐的新策略，最终在精度损失小于 1% 的前提下实现了模型体积缩减 60%。\n\nAwesome-Pruning 通过系统化整理碎片化的剪枝资源，帮助开发者从盲目的试错中解脱，专注于高效落地轻量级深度学习模型。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002Fhe-y_Awesome-Pruning_834105cf.png","he-y","Yang He","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002Fhe-y_d25ebe2b.png","Ph.D. student at UTS\r\n","University of Technology Sydney","Sydney, Australia",null,"https:\u002F\u002Fhe-y.github.io","https:\u002F\u002Fgithub.com\u002Fhe-y",2491,332,"2026-04-05T14:52:54",1,"","未说明",{"notes":91,"python":89,"dependencies":92},"Awesome-Pruning 是一个神经网络剪枝相关资源和论文的精选列表（Curated List），并非一个可直接运行的软件工具或代码库，因此 README 中未包含具体的操作系统、硬件配置、Python 版本或依赖库等运行环境需求。该仓库主要提供论文链接及部分论文对应的独立代码库地址（通常基于 PyTorch），用户需参考具体论文项目的文档以获取相应的环境要求。",[],[13,54],[95,96,97,98],"pruning","model-compression","model-acceleration","awesome-list","2026-03-27T02:49:30.150509","2026-04-06T08:45:45.621513",[102,107,112,117,122,127,132,137],{"id":103,"question_zh":104,"answer_zh":105,"source_url":106},9583,"这个仓库还在持续更新吗？","是的，仓库仍在更新中。维护者已确认并更新了 NeurIPS 2020 及之后的论文列表。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F23",{"id":108,"question_zh":109,"answer_zh":110,"source_url":111},9584,"如何提交新的剪枝论文或代码链接到列表中？","用户可以通过创建 Issue 提供论文标题、会议信息、论文链接和代码链接。维护者会在核实后将其添加到列表中（例如：ICML 2021、ECCV 2020 Oral 等论文均通过此方式添加）。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F27",{"id":113,"question_zh":114,"answer_zh":115,"source_url":116},9585,"如果发现列表中的论文类型标签（如结构化\u002F非结构化）标注错误怎么办？","用户可以在 Issue 中指出具体的错误论文及正确的类型（例如指出某篇论文应为 Filter-level 或 Weight-level），维护者会立即修正标签。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F4",{"id":118,"question_zh":119,"answer_zh":120,"source_url":121},9586,"列表中的论文年份或会议信息有误如何反馈？","请直接提交 Issue 说明正确的年份或会议名称（例如将 2016 ICLR 更正为 2017 ICLR），维护者确认后会进行修改。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F7",{"id":123,"question_zh":124,"answer_zh":125,"source_url":126},9587,"某些论文的代码链接失效或不存在该如何处理？","用户应提交 Issue 提供正确的代码链接，或者指出该论文尚未发布代码。维护者会根据反馈更新链接或移除无效链接。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F9",{"id":128,"question_zh":129,"answer_zh":130,"source_url":131},9588,"是否收录非会议期刊论文或在审论文？","仓库主要收录顶级会议论文。对于在审期刊论文，建议先等待正式发表或录用后再提交，具体可先在 Issue 中询问维护者意见。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F26",{"id":133,"question_zh":134,"answer_zh":135,"source_url":136},9589,"如何补充论文的研讨会版本或相关实现细节？","可以提交 Issue 提供额外的研讨会论文链接（如 ICML Workshop）或具体的代码实现示例（如 SelecSLS-Pytorch 中的隐式稀疏性应用），维护者会将这些补充信息加入列表。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F5",{"id":138,"question_zh":139,"answer_zh":140,"source_url":141},9590,"新发布的论文代码在哪里可以找到？","许多作者会在论文被接收后通过 Issue 通知维护者代码已开源（如 SCOP NeurIPS 2020, Only Train Once 等），列表中会随之更新具体的 GitHub 仓库链接。","https:\u002F\u002Fgithub.com\u002Fhe-y\u002FAwesome-Pruning\u002Fissues\u002F24",[]]