1 / 35

CS 5368: Artificial Intelligence Fall 2010

CS 5368: Artificial Intelligence Fall 2010. Lecture 12: MDP + RL (Part 2) 10/14/2010. Mohan Sridharan Slides adapted from Dan Klein. Recap: Reinforcement Learning. Basic idea: Receive feedback in the form of rewards. Agent’s utility is defined by the reward function.

miriam
Download Presentation

CS 5368: Artificial Intelligence Fall 2010

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. CS 5368: Artificial IntelligenceFall 2010 Lecture 12: MDP + RL (Part 2) 10/14/2010 Mohan Sridharan Slides adapted from Dan Klein

  2. Recap: Reinforcement Learning • Basic idea: • Receive feedback in the form of rewards. • Agent’s utility is defined by the reward function. • Must learn to act so as to maximize expected rewards.

  3. Reinforcement Learning Problem: Find values for fixed policy  (policy evaluation): Model-based learning: Learn the model, solve for values. Model-free learning: Solve for values directly (by sampling).

  4. The Story So Far: MDPs and RL Techniques: Things we know how to do: • If we know the MDP • Compute V*, Q*, * exactly. • Evaluate a fixed policy . • If we don’t know the MDP: • We can estimate the MDP then solve. • We can estimate V for a fixed policy . • We can estimate Q*(s,a) for the optimal policy while executing an exploration policy. • Model-based RL: • Value and policy Iteration. • Policy evaluation.. • Model-free RL: • Value learning. • Q-learning.

  5. Passive Learning • Simplified task: • You do not know the transitions T(s,a,s’). • You do not know the rewards R(s,a,s’). • You are given a policy (s). • Goal: learn the state values (and the model?). Policy evaluation. • In this case: • Learner “along for the ride”. • No choice about what actions to take. Just execute the policy and learn from experience. • This is NOT offline planning!

  6. Active Learning • More complex task. • R and T still unknown. Does not have fixed policy that determines its behavior! • Must learn what actions to take. • Same set of algorithms can be modified to address the challenges. • We begin with passive model-based and model-free methods.

  7. Example: Direct Estimation y • Episodes: +100 (1,1) up -1 (1,2) up -1 (1,2) up -1 (1,3) right -1 (2,3) right -1 (3,3) right -1 (3,2) up -1 (3,3) right -1 (4,3) exit +100 (done) (1,1) up -1 (1,2) up -1 (1,3) right -1 (2,3) right -1 (3,3) right -1 (3,2) up -1 (4,2) exit -100 (done) -100 x  = 1, R = -1 V(2,3) ~ (96 + -103) / 2 = -3.5 V(3,3) ~ (99 + 97 + -102) / 3 = 31.3

  8. Model-Based Learning • Idea: • Learn the model empirically through experience. • Solve for values as if the learned model were correct. • Simple empirical model learning: • Count outcomes for each s, a. • Normalize to give estimate of T(s, a, s’). • Discover R(s, a, s’) when we experience (s, a, s’). • Solving the MDP with the learned model: • Iterative policy evaluation, for example: s (s) s, (s) s, (s),s’ s’

  9. Example: Model-Based Learning y • Episodes: +100 (1,1) up -1 (1,2) up -1 (1,2) up -1 (1,3) right -1 (2,3) right -1 (3,3) right -1 (3,2) up -1 (3,3) right -1 (4,3) exit +100 (done) (1,1) up -1 (1,2) up -1 (1,3) right -1 (2,3) right -1 (3,3) right -1 (3,2) up -1 (4,2) exit -100 (done) -100 x  = 1 T(<3,3>, right, <4,3>) = 1 / 3 T(<2,3>, right, <3,3>) = 2 / 2

  10. Model-Free Learning • Want to compute an expectation weighted by P(x): • Model-based: estimate P(x) from samples, compute expectation. • Model-free: estimate expectation directly from samples. • Why does this work? Because samples appear with the right frequencies!

  11. Sample-Based Policy Evaluation? • Who needs T and R? Approximate the expectation with samples (drawn from T). s (s) s, (s),s’ s, (s) s’ s1’ s3’ s2’ Almost! But we only actually make progress when we move to i+1.

  12. Temporal-Difference Learning • Big idea: learn from every experience! • Update V(s) each time we experience (s,a,s’,r) • Likely s’ will contribute to updates more often. • Temporal difference learning: • Policy can still be fixed! • Move values toward value of whatever successor occurs: running average! s (s) s, (s) s’ Sample of V(s): Update to V(s): Same update:

  13. Example: TD Policy Evaluation (1,1) up -1 (1,2) up -1 (1,2) up -1 (1,3) right -1 (2,3) right -1 (3,3) right -1 (3,2) up -1 (3,3) right -1 (4,3) exit +100 (done) (1,1) up -1 (1,2) up -1 (1,3) right -1 (2,3) right -1 (3,3) right -1 (3,2) up -1 (4,2) exit -100 (done) Take  = 1,  = 0.5

  14. s a s, a s,a,s’ s’ Problems with TD Value Learning • TD value leaning is a model-free way to do policy evaluation. • However, if we want to turn values into a (new) policy, we are sunk: • Idea: learn Q-values directly. • Makes action selection model-free too!

  15. Return to Active Learning • Full reinforcement learning • You do not know the transitions T(s,a,s’) • You do not know the rewards R(s,a,s’) • You can choose any actions you like. • Goal: learn the optimal policy. • … what value iteration did! • In this case: • Learner makes choices! • Fundamental tradeoff: exploration vs. exploitation. • You actually take actions in the world and find out what happens.

  16. Model-Based Active Learning • In general, want to learn the optimal policy, not evaluate a fixed policy. • Idea: adaptive dynamic programming  • Learn an initial model of the environment. • Solve for the optimal policy for this model (value or policy iteration). • Refine model through experience and repeat. • Crucial: we have to make sure we actually learn about all of the model.

  17. Example: Greedy ADP • Imagine we find the lower path to the good exit first. • Some states will never be visited following this policy from (1,1). • Can keep re-using this policy but following it never explores the regions of the model we need in order to learn the optimal policy . ? ?

  18. What Went Wrong? • Problem with following optimal policy for current model: • Never learns about better regions of the space if current policy neglects them. • Fundamental tradeoff: exploration vs. exploitation. • Exploration: take actions with suboptimal estimates to discover new rewards and increase eventual utility. • Exploitation: once the true optimal policy is learned, exploration reduces utility. • Systems must explore in the beginning and exploit in the limit. ? ?

  19. Detour: Q-Value Iteration • Value iteration: find successive approx optimal values • Start with V0*(s) = 0, which we know is right (why?) • Given Vi*, calculate the values for all states for depth i+1: • But Q-values are more useful! • Start with Q0*(s,a) = 0, which we know is right (why?) • Given Qi*, calculate the q-values for all q-states for depth i+1:

  20. Q-Learning • We would like to do Q-value updates to each Q-state: • But cannot compute this update without knowing T, R. • Instead, compute average as we go: • Receive a sample transition (s,a,r,s’). • This sample suggests: • But we want to average over results from (s,a) (Why?) • So keep a running average:

  21. Q-Learning Properties • Will converge to optimal policy: • If you explore enough (i.e. visit each q-state many times). • If you make the learning rate small enough. • Basically does not matter how you select actions (!) • Off-policy learning: learns optimal q-values, not the values of the policy you are following. • On-policy vs. off-policy: Chapter 5 on RL textbook.

  22. Exploration / Exploitation • Several schemes for forcing exploration: • Simplest: random actions ( greedy). • Every time step, flip a coin. • With probability , act randomly. • With probability 1-, act according to current policy. • Regret: expected gap between rewards during learning and rewards from optimal action. • Q-learning with random actions will converge to optimal values, but possibly very slowly, and will get low rewards on the way. • Results will be optimal but regret will be large. • How to make regret small?

  23. Exploration Functions • When to explore: • Random actions: explore a fixed amount • Better ideas: explore areas whose badness is not (yet) established, explore less over time • One way: exploration function. • Takes a value estimate and a count, and returns an optimistic utility, (exact form not important):

  24. Q-Learning • Q-learning produces tables of q-values:

  25. Q-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 will see it over and over again!

  26. Example: Pacman • Let’s say we discover through experience that this state is bad. • In naïve Q-learning, we know nothing about this state or its q-states. • Or even this one!

  27. Feature-Based Representations • Solution: describe a state using a vector of features (properties). • Features map from states to real numbers that capture important properties of the state. • Example features: • Distance to closest ghost/dot. • Number of ghosts. • 1 / (dist to dot)2 • Is Pacman in a tunnel? (0/1) • 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).

  28. Linear Feature Functions • Using a feature representation, 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 actually be very different in value 

  29. Function Approximation • Q-learning with linear q-functions: • Intuitive interpretation: • Adjust weights of active features. • E.g. if something unexpectedly bad happens, do not prefer all states with that state’s features. • Formal justification: online least squares. Exact Q’s Approximate Q’s

  30. Example: Q-Pacman

  31. 26 24 22 20 30 40 20 30 20 10 10 0 0 Linear Regression 40 20 0 0 20 Prediction Prediction

  32. Overfitting 30 25 Degree 15 polynomial 20 15 10 5 0 -5 -10 -15 0 2 4 6 8 10 12 14 16 18 20

  33. Policy Search

  34. Policy Search • Problem: often the feature-based policies that work well are not the ones that approximate V or Q best. • E.g. value functions may provide horrible estimates of future rewards, but they can still produce good decisions. • Will see 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 helicopter.

  35. Policy 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 impractical 

More Related