Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Noisy Networks for Exploration
(Submitted on 30 Jun 2017 (v1), last revised 9 Jul 2019 (this version, v3))
Abstract: We introduce NoisyNet, a deep reinforcement learning agent with parametric noise added to its weights, and show that the induced stochasticity of the agent's policy can be used to aid efficient exploration. The parameters of the noise are learned with gradient descent along with the remaining network weights. NoisyNet is straightforward to implement and adds little computational overhead. We find that replacing the conventional exploration heuristics for A3C, DQN and dueling agents (entropy reward and $\epsilon$-greedy respectively) with NoisyNet yields substantially higher scores for a wide range of Atari games, in some cases advancing the agent from sub to super-human performance.
Submission history
From: Charles Blundell [view email][v1] Fri, 30 Jun 2017 17:56:19 GMT (5889kb,D)
[v2] Thu, 15 Feb 2018 16:00:54 GMT (5167kb,D)
[v3] Tue, 9 Jul 2019 09:57:23 GMT (5167kb,D)
Link back to: arXiv, form interface, contact.