rl.tutorial

Outline of Tutorial1. IntroductionAn Introduction to Reinforcement 2. Elements of Reinforcement Learning3. Solution methodsLearning4. Survey of other subjectsTim KovacsUniversity of Bristolkovacs@cs.bris.ac.uk21 What is it? 1 Points of ViewThe learning agent’s point of view:Reinforcement Learning is learning how to act in order to maximise a • RL is learning from trial and error interaction with the world.numerical reward signal. • E.g. how much reward will I get if I do this?RL is:• a class of tasks “RL as a tool” point of view:• which require a kind of trial-and-error learning • RL is training by rewards and punishments.• Train the computer as we might train a dog.Features of RL:• learning from numerical rewards Application areas: problems with ongoing interaction e.g.• interaction with the task; sequences of states, actions and rewards • robotics• uncertainty and non-deterministic worlds • animal learning• delayed consequences • scheduling• the explore/exploit dilemma • games• the whole problem of goal-directed learning • control systems3 41 Supervised Learning 1 Reinforcement LearningStep: 1 Step: 1Teacher: Does picture 1 show a car or a flower? World: You are in state 9. Choose action A or C.Learner: A flower. Learner: Action A.Teacher: No, it’s a car. World: Your reward is 100.Step: 2 Step: 2Teacher: Does picture 2 show a car or a flower? World: You are in state 32. Choose action B or E.Learner: A car. Learner: Action B.Teacher: Yes ...