Policy Gradient Critics Export

@incollection{citeulike:2361554, abstract = {We present Policy Gradient Actor-Critic (PGAC), a new model-free Reinforcement Learning (RL) method for creating limited-memory stochastic policies for Partially Observable Markov Decision Processes (POMDPs) that require long-term memories of past observations and actions. The approach involves estimating a policy gradient for an Actor through a Policy Gradient Critic which evaluates probability distributions on actions. Gradient-based updates of history-conditional action probability distributions enable the algorithm to learn a mapping from memory states (or event histories) to probability distributions on actions, solving POMDPs through a combination of memory and stochasticity. This goes beyond previous approaches to learning purely reactive POMDP policies, without giving up their advantages. Preliminary results on important benchmark tasks show that our approach can in principle be used as a general purpose POMDP algorithm that solves RL problems in both continuous and discrete action domains.}, author = {Wierstra, Daan and Schmidhuber, J\"{u}rgen}, citeulike-article-id = {2361554}, citeulike-linkout-0 = {http://dx.doi.org/10.1007/978-3-540-74958-5_43}, doi = {10.1007/978-3-540-74958-5_43}, journal = {Machine Learning: ECML 2007}, keywords = {action\_selection, actor-critic, policy\_gradient}, pages = {466--477}, posted-at = {2008-02-11 07:11:11}, priority = {0}, title = {Policy Gradient Critics}, url = {http://dx.doi.org/10.1007/978-3-540-74958-5_43}, year = {2007} }

TY - CHAP ID - citeulike:2361554 L3 - citeulike-article-id:2361554 N2 - We present Policy Gradient Actor-Critic (PGAC), a new model-free Reinforcement Learning (RL) method for creating limited-memory stochastic policies for Partially Observable Markov Decision Processes (POMDPs) that require long-term memories of past observations and actions. The approach involves estimating a policy gradient for an Actor through a Policy Gradient Critic which evaluates probability distributions on actions. Gradient-based updates of history-conditional action probability distributions enable the algorithm to learn a mapping from memory states (or event histories) to probability distributions on actions, solving POMDPs through a combination of memory and stochasticity. This goes beyond previous approaches to learning purely reactive POMDP policies, without giving up their advantages. Preliminary results on important benchmark tasks show that our approach can in principle be used as a general purpose POMDP algorithm that solves RL problems in both continuous and discrete action domains. JF - Machine Learning: ECML 2007 EP - 477 TI - Policy Gradient Critics SP - 466 KW - action_selection KW - actor-critic KW - policy_gradient AU - Wierstra, Daan AU - Schmidhuber, Jürgen PY - 2007/// UR - http://dx.doi.org/10.1007/978-3-540-74958-5_43 ER -