reinforcement learning课件.ppt

Reinforcement Learning;Previous Lectures;Today;Robot in a room;Other examples;Outline;Robot in a room;Is this a solution?;Optimal policy;Reward for each step -2;Reward for each step: -0.1;Reward for each step: -0.04;Reward for each step: -0.01;Reward for each step: +0.01;Markov Decision Process (MDP);Computing return from rewards;Value functions;Optimal value functions;Outline;Dynamic programming;Policy evaluation/improvement;Policy/Value iteration;Using DP;Outline;Monte Carlo methods;Monte Carlo policy evaluation;Monte Carlo control;Maintaining exploration;Simulated experience;Summary of Monte Carlo;Outline;Temporal Difference Learning;MC vs. TD;Sarsa;Q-learning;Outline;State representation;Function approximation;Features;Splitting and aggregation;Designing rewards;Case study: Back gammon;Summary;www.cs.ualberta.ca/~sutton/book/the-book.html