140篇文章 · 218074字 · 183人關(guān)注
-
Concrete dropout
一種采用貝葉斯學(xué)習(xí)的dropoout方法變體 Dropout is used as a practical tool to obtain unc...
-
深度學(xué)習(xí)硬件——一切待定的未來在 Titan V 發(fā)布后，我們進(jìn)入了深度學(xué)習(xí)硬件的待定狀態(tài)。NVIDIA 深度學(xué)習(xí)硬件的賣主是否被替代還不明晰。AMD 和 Intel Ner...
-
Cortical microcircuits as gated-recurrent neural networks
Rui Ponte Costa?Centre for Neural Circuits and BehaviourDept. of Physiol...
-
HIERARCHICAL REPRESENTATIONS FOR EFFICIENT ARCHITECTURE SEARCH
Hanxiao Liu?Carnegie Mellon Universityhanxiaol@cs.cmu.eduKaren Simonyan,...
-
Knowledge Sharing for Reinforcement Learning: Writing a BOOK
Knowledge Sharing for Reinforcement Learning:Writing a BOOKSimyung Chang...
-
Emergent Complexity via Multi-Agent Competition
Trapit Bansal?UMass AmherstJakub PachockiOpenAISzymon SidorOpenAIIlya Su...
-
Inverse Reward Design
Dylan Hadfield-Menell Smitha Milli Pieter Abbeel? Stuart Russell Anca Dr...
-
CAN DEEP REINFORCEMENT LEARNING SOLVE ERDOS-SELFRIDGE-SPENCER GAMES?
Maithra RaghuGoogle Brain and Cornell University{maithrar}@gmail.comAlex...
-
Neural Discrete Representation Learning
Learning useful representations without supervision remains a key challe...