Model-Free Episodic Control_DeepMind英文文献资料.pdf

Model-Free Episodic Control Charles Blundell Benigno Uria Alexander Pritzel Google DeepMind Google DeepMind Google DeepMind cblundell@ buria@ apritzel@ 6 Yazhe Li Avraham Ruderman Joel Z Leibo Jack Rae 1 Google DeepMind Google DeepMind Google DeepMind Google DeepMind 0 yazhe@ aruderman@ jzl@ jwrae@ 2 n u Daan Wierstra Demis Hassabis J Google DeepMind Google DeepMind 4 wierstra@ demishassabis@ 1 ] L Abstract M. State of the art deep reinforcement learning algorithms take many millions of inter- t actions to attain human-level performance. Humans, on the other hand, can very a t quickly exploit highly rewarding nuances of an environment upon first discovery. s In the brain, such rapid learning is thought to depend on the hippocampus and [ its capacity for episodic memory. Here we investigate whether a simple model 1 of hippocampal episodic control can learn to solve difficult sequential decision- v making tasks. We demonstrate that it not only attains a highly rewarding strategy 0 significantly faster than state-of-the-art deep reinforcement learning algorithms, but 6 also achieves a higher overall reward on some of the more challenging do