Parameter Space Noise for Exploration

Deep reinforcement learning (RL) methods generally engage in exploratory behavior through noise injection in the action space. An alternative is to add noise directly to the agent's parameters, which can lead to more consistent exploration and a richer set of behaviors. Methods such as evolutionary strategies use parameter perturbations, but discard all temporal structure in the process and require significantly more samples. Combining parameter noise with traditional RL methods allows to combine the best of both worlds. We demonstrate that both off- and on-policy methods benefit from this approach through experimental comparison of DQN, DDPG, and TRPO on high-dimensional discrete action environments as well as continuous control tasks. Our results show that RL with parameter noise learns more efficiently than traditional RL with action space noise and evolutionary strategies individually.

[1]  G. Uhlenbeck,et al.  On the Theory of the Brownian Motion , 1930 .

[2]  Ingo Rechenberg,et al.  Evolutionsstrategie : Optimierung technischer Systeme nach Prinzipien der biologischen Evolution , 1973 .

[3]  H. P. Schwefel,et al.  Numerische Optimierung von Computermodellen mittels der Evo-lutionsstrategie , 1977 .

[4]  Sebastian Thrun,et al.  Efficient Exploration In Reinforcement Learning , 1992 .

[5]  Richard S. Sutton,et al.  Introduction to Reinforcement Learning , 1998 .

[6]  Sham M. Kakade,et al.  A Natural Policy Gradient , 2001, NIPS.

[7]  Ronen I. Brafman,et al.  R-MAX - A General Polynomial Time Algorithm for Near-Optimal Reinforcement Learning , 2001, J. Mach. Learn. Res..

[8]  Ananth Ranganathan,et al.  The Levenberg-Marquardt Algorithm , 2004 .

[9]  Ronald J. Williams,et al.  Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning , 2004, Machine Learning.

[10]  Michael Kearns,et al.  Near-Optimal Reinforcement Learning in Polynomial Time , 1998, Machine Learning.

[11]  Richard S. Sutton,et al.  Reinforcement Learning: An Introduction , 1998, IEEE Trans. Neural Networks.

[12]  Stefan Schaal,et al.  Natural Actor-Critic , 2003, Neurocomputing.

[13]  Peter Auer,et al.  Near-optimal Regret Bounds for Reinforcement Learning , 2008, J. Mach. Learn. Res..

[14]  Jürgen Schmidhuber,et al.  State-Dependent Exploration for Policy Gradient Methods , 2008, ECML/PKDD.

[15]  Tom Schaul,et al.  Natural Evolution Strategies , 2008, 2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence).

[16]  Jan Peters,et al.  Policy Search for Motor Primitives in Robotics , 2008, NIPS 2008.

[17]  Tom Schaul,et al.  Efficient natural evolution strategies , 2009, GECCO.

[18]  Tom Schaul,et al.  Stochastic search using the natural gradient , 2009, ICML '09.

[19]  Tom Schaul,et al.  Exponential natural evolution strategies , 2010, GECCO '10.

[20]  Hado van Hasselt,et al.  Double Q-learning , 2010, NIPS.

[21]  Tom Schaul,et al.  A Natural Evolution Strategy for Multi-objective Optimization , 2010, PPSN.

[22]  Frank Sehnke,et al.  Parameter-exploring policy gradients , 2010, Neural Networks.

[23]  Eduardo F. Morales,et al.  An Introduction to Reinforcement Learning , 2011 .

[24]  Tom Schaul,et al.  High dimensions and heavy tails for natural evolution strategies , 2011, GECCO '11.

[25]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[26]  Sergey Levine,et al.  Incentivizing Exploration In Reinforcement Learning With Deep Predictive Models , 2015, ArXiv.

[27]  Sergey Levine,et al.  Trust Region Policy Optimization , 2015, ICML.

[28]  Jimmy Ba,et al.  Adam: A Method for Stochastic Optimization , 2014, ICLR.

[29]  Shane Legg,et al.  Human-level control through deep reinforcement learning , 2015, Nature.

[30]  Marc G. Bellemare,et al.  The Arcade Learning Environment: An Evaluation Platform for General Agents (Extended Abstract) , 2012, IJCAI.

[31]  Yuval Tassa,et al.  Continuous control with deep reinforcement learning , 2015, ICLR.

[32]  Benjamin Van Roy,et al.  Generalization and Exploration via Randomized Value Functions , 2014, ICML.

[33]  Pieter Abbeel,et al.  Benchmarking Deep Reinforcement Learning for Continuous Control , 2016, ICML.

[34]  Filip De Turck,et al.  VIME: Variational Information Maximizing Exploration , 2016, NIPS.

[35]  Benjamin Van Roy,et al.  Deep Exploration via Bootstrapped DQN , 2016, NIPS.

[36]  Tom Schaul,et al.  Dueling Network Architectures for Deep Reinforcement Learning , 2015, ICML.

[37]  Tom Schaul,et al.  Unifying Count-Based Exploration and Intrinsic Motivation , 2016, NIPS.

[38]  Geoffrey E. Hinton,et al.  Layer Normalization , 2016, ArXiv.

[39]  Tom Schaul,et al.  Prioritized Experience Replay , 2015, ICLR.

[40]  Zoubin Ghahramani,et al.  Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning , 2015, ICML.

[41]  Wojciech Zaremba,et al.  OpenAI Gym , 2016, ArXiv.

[42]  Filip De Turck,et al.  #Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning , 2016, NIPS.

[43]  Alexei A. Efros,et al.  Curiosity-Driven Exploration by Self-Supervised Prediction , 2017, 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[44]  Shipra Agrawal,et al.  A Near-optimal Regret Bounds for , 2017 .

[45]  Xi Chen,et al.  Evolution Strategies as a Scalable Alternative to Reinforcement Learning , 2017, ArXiv.

[46]  A. P. Hyper-parameters Count-Based Exploration with Neural Density Models , 2017 .

[47]  S. Shankar Sastry,et al.  Surprise-Based Intrinsic Motivation for Deep Reinforcement Learning , 2017, ArXiv.

[48]  Shane Legg,et al.  Noisy Networks for Exploration , 2017, ICLR.

[49]  Ilya Kostrikov,et al.  Intrinsic Motivation and Automatic Curricula via Asymmetric Self-Play , 2017, ICLR.