1CS 188: Artificial IntelligenceFall 2008Lecture 12: Reinforcement LearningLecture 12: Reinforcement Learning10/7/2008Dan Klein – UC BerkeleyMany slides over the course adapted from either Stuart Russell or Andrew Moore12Reinforcement Learning Reinforcement learning: Still have an MDP: A set of states s ∈ S A set of actions (per state) AA model T(s,a,s’)A model T(s,a,s’) A reward function R(s,a,s’) Still looking for a policy π(s) New twist: don’t know T or R I.e. don’t know which states are good or what the actions do Must actually try actions and states out to learn[DEMO]33Model-Free Learning Temporal difference learning Update each time we experience a transition Frequent outcomes will contribute more updates (over time)π(s)ss, π(s)4s’4Q-Learning Learn Q*(s,a) values Receive a sample (s,a,s’,r) Consider your old estimate: Consider your new sample estimate:[DEMO – Grid Q’s] Incorporate the new estimate into a running average:55Q-Learning Properties Will converge to optimal policy If you explore enough If you make the learning rate small enough But not decrease it too quickly!Basically doesn’t matter how you select actions (!)[DEMO – Grid Q’s]Basically doesn’t matter how you select actions (!) Neat property: learns optimal q-values regardless of action selection noise (some caveats)S ES E66Exploration / Exploitation Several schemes for forcing exploration Simplest: random actions (ε greedy) Every time step, flip a coin With probability ε, act randomlyWith probability 1-ε, act according to current policy[DEMO – RL Pacman]With probability 1-ε, act according to current policy Problems with random actions? You do explore the space, but keep thrashing around once learning is done One solution: lower ε over time Another solution: exploration functions77Exploration Functions When to explore Random actions: explore a fixed amount Better idea: explore areas whose badness is not (yet) established Exploration function Takes a value estimate and a count, and returns an optimistic utility, e.g. (exact form not important)88Q-Learning Q-learning produces tables of q-values:[DEMO – Crawler Q’s]99Q-Learning In realistic situations, we cannot possibly learn about every single state! Too many states to visit them all in training Too many states to hold the q-tables in memory Instead, we want to generalize: Learn about some small number of training states from experience Generalize that experience to new, similar states This is a fundamental idea in machine learning, and we’ll see it over and over again1010Example: Pacman Let’s say we discover through experience that this state is bad:In naïve q learning, we In naïve q learning, we know nothing about this state or its q states: Or even this one!1111Feature-Based Representations Solution: describe a state using a vector of features Features are functions from states to real numbers (often 0/1) that capture important properties of the stateExample features:Example features: Distance to closest ghost Distance to closest dot Number of ghosts 1 / (dist to dot)2 Is Pacman in a tunnel? (0/1) …… etc. Is it the exact state on this slide? Can also describe a q-state (s, a) with features (e.g. action moves closer to food)1212Linear Feature Functions Using a feature representation, we can write a q function (or value function) for any state using a few weights: Advantage: our experience is summed up in a few powerful numbers Disadvantage: states may share features but be very different in value!1313Function Approximation Q-learning with linear q-functions: Intuitive interpretation: Adjust weights of active features E.g. if something unexpectedly bad happens, disprefer all states with that state’s features Formal justification: online least squares1414Example: Q-Pacman1515Linear Regression2022242620400102030400102030200 10 200Given examplesPredictgiven a new point1616204020222426Linear Regression0 200010203040010203020PredictionPrediction1717Ordinary Least Squares (OLS)Error or “residual”Observation0 200Prediction1818Minimizing ErrorApproximate q update explained:19191015202530Degree 15 polynomialOverfitting0 2 4 6 8 10 12 14 16 18 20-15-10-505[DEMO]20Policy Search [DEMO – Helicopter]2121Policy Search2222Policy Search Problem: often the feature-based policies that work well aren’t the ones that approximate V / Q best E.g. your value functions from project 2 were probably horrible estimates of future rewards, but they still produced good decisionsWe’ll see this distinction between modeling and prediction again We’ll see this distinction between modeling and prediction again later in the course Solution: learn the policy that maximizes rewards rather than the value that predicts rewards This is the idea behind policy search, such as what controlled the upside-down helicopter2323Policy Search Simplest policy search: Start with an initial linear value function or q-function Nudge each feature weight up and down and see if your policy is better than before Problems: How do we tell the policy got better? Need to run many sample episodes! If there are a lot of features, this can be impractical2424Policy Search* Advanced policy search: Write a stochastic (soft) policy: Turns out you can efficiently approximate the derivative of the returns with respect to the parameters w (details in the book, optional material) Take uphill steps, recalculate derivatives, etc.2525Take a Deep Breath… We’re done with search and planning! Next, we’ll look at how to reason with probabilities Diagnosis Tracking objects Speech recognition Robot mapping … lots more! Last part of course: machine
View Full Document