【导读】今天给大家整理了CVPR2020录用的几篇神经网络架构搜索方面的论文,神经网络架构搜索又称为Neural Architecture Search,简称(NAS)。神经网络架构搜索在这两年比较热门,学术界和国内外知名企业都在做这方面的研究。之后,本公众号后续将出一个NAS方面的专辑,主要包括NAS的发展历程、论文解读和应用场景。希望大家多多关注
!
论文汇总
1.Blockwisely Supervised Neural Architecture Search with Knowledge Distillation(该论文在ImageNet数据集进行训练得到了78.4% top-1 accuracy ,比EfficientNet-B0高了2.1%个点)
- 作者团队:暗物智能、Monash 大学、中山大学
- 论文链接:https://arxiv.org/abs/1911.13053
2. Semi-Supervised Neural Architecture Search
- 作者团队:MSRA、中科大
- 论文链接:https://arxiv.org/abs/2002.10389
- 代码地址:https://github.com/renqianluo/SemiNAS
3. CARS: Continuous Evolution for Efficient Neural Architecture Search
- 作者团队:北大、华为诺亚、鹏城实验室、悉尼大学
- 论文链接:https://arxiv.org/abs/1909.04977
- 代码(即将开源):https://github.com/huawei-noah/CARS
4. Densely Connected Search Space for More Flexible Neural Architecture Search
- 论文链接:https://arxiv.org/abs/1906.09607
- 代码地址:https://github.com/JaminFong/DenseNAS
5. AdversarialNAS: Adversarial Neural Architecture Search for GANs
- 论文链接:https://arxiv.org/pdf/1912.02037.pdf
- 代码地址:https://github.com/chengaopro/AdversarialNAS
6. Hit-Detector: Hierarchical Trinity Architecture Search for Object Detection
- 作者团队:北大、华为诺亚、悉尼大学
- 论文链接:https://arxiv.org/pdf/2003.11818.pdf
- 代码地址:https://github.com/ggjy/HitDet.pytorch
7. AOWS: Adaptive and optimal network width search with latency constraints
- 论文链接:https://arxiv.org/abs/2005.10481
- 代码地址:https://github.com/bermanmaxim/AOWS
8. MTL-NAS: Task-Agnostic Neural Architecture Search towards General-Purpose Multi-Task Learning
- 论文:https://arxiv.org/abs/2003.14058
- 代码:https://github.com/bhpfelix/MTLNAS
9. Neural Architecture Search for Lightweight Non-Local Networks
- 论文:https://arxiv.org/abs/2004.01961
- 代码:https://github.com/LiYingwei/AutoNL
10. SGAS: Sequential Greedy Architecture Search
- 作者团队:KAUST, Intel
- 论文链接:https://arxiv.org/pdf/1912.00195.pdf
- 代码地址:https://www.deepgcns.org/auto/sgas
11. GreedyNAS: Towards Fast One-Shot NAS with Greedy Supernet
- 作者团队:商汤、清华、Dian、华科
- 论文链接:https://arxiv.org/abs/2003.11236
12. FBNetV2: Differentiable Neural Architecture Search for Spatial and Channel Dimensions(UC Berkley, Facebook)
- 论文链接:https://arxiv.org/abs/2004.05565
- 代码地址:https://github.com/facebookresearch/mobile-vision
13. MiLeNAS: Efficient Neural Architecture Search via Mixed-Level Reformulation
- 作者团队:南加州、腾讯、港中文、港科大
- 论文链接:https://arxiv.org/abs/2003.12238
- 代码地址:https://github.com/chaoyanghe/MiLeNAS
14. Designing Network Design Spaces
- 作者团队:Facebook FAIR(何凯明团队)
- 论文链接:https://arxiv.org/abs/2003.13678
15. Search to Distill: Pearls are Everywhere but not the Eyes
- 作者团队:Google,港中文
- 论文链接:https://arxiv.org/abs/1911.09074
16. EcoNAS: Finding Proxies for Economical Neural Architecture Search
- 作者团队:悉尼大学,南洋理工,商汤
- 论文链接:https://arxiv.org/abs/2001.01233
17.DSNAS: Direct Neural Architecture Search without Parameter Retraining
- 作者团队:港中文、UCLA、剑桥、商汤
- 论文链接:https://arxiv.org/abs/2002.09128
18.MobileDets: Searching for Object Detection Architectures for Mobile Accelerators
- 论文作者:谷歌、威斯康星大学麦迪逊分校
- 论文链接:https://arxiv.org/abs/2004.14525
19. Rethinking Performance Estimation in Neural Architecture Search
- 论文:https://arxiv.org/abs/2005.09917
- 代码:https://github.com/zhengxiawu/rethinking_performance_estimation_in_NAS
- 解读1:https://www.zhihu.com/question/372070853/answer/1035234510
- 解读2:https://zhuanlan.zhihu.com/p/111167409
20. When NAS Meets Robustness: InSearchof RobustArchitecturesagainst Adversarial Attacks
- 作者团队:港中文、 MIT
- 论文链接:https://arxiv.org/abs/1911.10695
- 代码地址:https://github.com/gmh14/RobNets