本週的重要論文有 DeepMind 聯合哈佛大學等機構造出 3D 模擬小老鼠來探索神經網路,以及伯克利提出的資料增強加持的強化學習方法實現了多環境下的 SOTA結果。
DeepEEP Neuroethology of A Virtual Rodent
Training with Quantization Noise for Extreme Model Compression
Lite Transformer With Long-short Range Attention
Decoupling Representation and classifier for long-tailed recognition
Bayesian Deep Learning and a Probabilistic Perspective of Generalization
MakeItTalk: Speaker-Aware Talking Head Animation
Reinforcement Learning with Augmented Data
ArXiv Weekly Radiostation:NLP、CV、ML更多精選論文(附音訊)
作者:Josh Merel、Diego Aldarondo、Bence Olveczky 等
論文連結:https://openreview.net/pdf?id=SyxrxR4KPS
作者:Angela Fan、Pierre Stock、Armand Joulin 等
論文連結:https://arxiv.org/pdf/2004.07320.pdf
作者:Zhanghao Wu、Zhijian Liu、Song Han
論文連結:https://arxiv.org/pdf/2004.11886v1.pdf
作者:Bingyi Kang、Saining Xie、Yannis Kalantidis 等
論文連結:https://openreview.net/pdf?id=r1gRTCVFvB
作者:Andrew Gordon Wilson、Pavel Izmailov
論文連結:https://arxiv.org/pdf/2002.08791.pdf
作者:Yang Zhou, DIngzeyu Li、Jose Echevarria 等
論文連結:https://arxiv.org/pdf/2004.12992v1.pdf
作者:Michael Laskin、Kimin Lee、Aravind Srinivas 等
論文連結:https://arxiv.org/pdf/2004.14990.pdf