Tatsuya Matsushima

icon indicating a website link t-matsushima.com icon indicating an email [email protected]

icon company The University of Tokyo @matsuolab @matsuolab-research icon location Tokyo, Japan icon location Postdoc Researcher in School of Engineering, The University of Tokyo

Results 59 comments of Tatsuya Matsushima

[Feature Request] Reducing memory consumption when using HerReplayBuffer

oh, sorry I overlooked this issue. thanks!

Reptile: a Scalable Metalearning Algorithm

openAI blog - https://blog.openai.com/reptile/ - 実装例あり github - https://github.com/openai/supervised-reptile

Sim-to-Real Transfer of Robotic Control with Dynamics Randomization

実機でzeroshot(初めて？)

Learning to Act by Predicting the Future

https://www.slideshare.net/DeepLearningJP2016/dllearning-to-act-by-predicting-the-future にまとめた

Towards deep learning with segregated dendrites

github - https://github.com/jordan-g/Segregated-Dendrite-Deep-Learning

Reinforcement Learning with Unsupervised Auxiliary Tasks

arXivTImes - https://github.com/arXivTimes/arXivTimes/issues/56 DeepMind blog - https://deepmind.com/blog/reinforcement-learning-unsupervised-auxiliary-tasks/

Reinforcement Learning with Unsupervised Auxiliary Tasks

UNREAL

Reinforcement Learning with Unsupervised Auxiliary Tasks

cited in #178

Adaptive Computation Time for Recurrent Neural Networks

> This paper introduces Adaptive Computation Time (ACT), an algorithm that allows recurrent neural networks to learn how many computational steps to take between receiving an input and emitting an...

Adaptive Computation Time for Recurrent Neural Networks

cited in #181