Sp14 Cs188 Lecture 9 - Mdps II

CS 188: Artificial Intelligence
Markov Decision Processes II
Instructors: Dan Klein and Pieter Abbeel --- University of California, Berkeley
[These slides were created by Dan Klein and Pieter Abbeel for CS188 Intro to AI at UC Berkeley. All CS188 materials are available at
Example: Grid World

A maze-like problem
The agent lives in a grid
Walls block the agents path
Noisy movement: actions do not always go as
planned
80% of the time, the action North takes
the agent North
10% of the time, North takes the agent
West; 10% East
If there is a wall in the direction the agent
would have been taken, the agent stays
put
The agent receives rewards each time step
Small living reward each step (can be
negative)
Recap: MDPs
Markov decision processes:
States S
Actions A
Transitions P(s|s,a) (or T(s,a,s))
Rewards R(s,a,s) (and discount )
Start state s0
s
a
s, a
s,a,s
s
Quantities:
Policy = map of states to actions

Utility = sum of discounted rewards
Values = expected future utility from a state (max node)
Q-Values = expected future utility from a q-state (chance node)
Optimal Quantities
The value (utility) of a state s:
V*(s) = expected utility starting in
s and acting optimally
s is a
state
The value (utility) of a q-state

(s,a):
Q*(s,a) = expected utility starting
out having taken action a from
state s and (thereafter) acting
optimally
The optimal policy:
*(s) = optimal action from state
(s, a) is a
q-state
s, a
s,a,s
s
(s,a,s) is a
transition
[Demo: gridworld values (L9D1)]
Gridworld Values V*
Gridworld: Q*
The Bellman Equations

How to be optimal:
Step 1: Take correct first
action
Step 2: Keep being optimal
The Bellman Equations

s
Definition of optimal utility via expectimax

recurrence gives a simple one-step
lookahead relationship amongst optimal
utility values
a
s, a
s,a,s
s
These are the Bellman equations, and they

characterize optimal values in a way well
Value Iteration
Bellman equations characterize the optimal values:
V(s)
a
s, a
s,a,s
Value iteration computes them:
Value iteration is just a fixed point solution method
though the Vk vectors are also interpretable as time-limited values
V(s)
Convergence*
How do we know the Vk vectors are going to
converge?
Case 1: If the tree has maximum depth M,
then VM holds the actual untruncated values
Case 2: If the discount is less than 1
Sketch: For any state Vk and Vk+1 can be
viewed as depth k+1 expectimax results in
nearly identical search trees
The difference is that on the bottom layer, Vk+1
has actual rewards while Vk has zeros
That last layer is at best all RMAX
It is at worst RMIN
But everything is discounted by k that far out
So Vk and Vk+1 are at most k max|R| different
Policy Methods
Policy Evaluation
Fixed Policies
Do the optimal action
Do what says to do
s
(s)
s, (s)
s, a
s,a,s
s
s,
(s),s
Expectimax trees max over all actions to compute the optimal values
If we fixed some policy (s), then the tree would be simpler only one
action per state
though the trees value would depend on which policy we fixed
Utilities for a Fixed Policy

Another basic operation: compute the utility of a
state s under a fixed (generally non-optimal)
policy
Define the utility of a state s, under a fixed policy
:
V(s) = expected total discounted rewards starting in s and
following
Recursive relation (one-step look-ahead / Bellman

equation):
s
(s)
s, (s)
s,
(s),s
Example: Policy Evaluation

Always Go Right
Always Go Forward
Example: Policy Evaluation

Always Go Right
Always Go Forward
Policy Evaluation
s
How do we calculate the Vs for a fixed policy ?
(s)
Idea 1: Turn recursive Bellman equations into updates

(like value iteration)
s, (s)
s,
(s),s
Efficiency: O(S2) per iteration

Idea 2: Without the maxes, the Bellman equations are just a linear
system
Solve with Matlab (or your favorite linear system solver)
Policy Extraction
Computing Actions from Values

Lets imagine we have the optimal values V*(s)
How should we act?
Its not obvious!
We need to do a mini-expectimax (one step)
This is called policy extraction, since it gets the policy

implied by the values
Computing Actions from Q-Values

Lets imagine we have the optimal q-values:
How should we act?
Completely trivial to decide!
Important lesson: actions are easier to select from q-values

than values!
Policy Iteration
Problems with Value Iteration

Value iteration repeats the Bellman updates:
s
a
s, a
Problem 1: Its slow O(S2A) per iteration
s,a,s
s
Problem 2: The max at each state rarely changes

Problem 3: The policy often converges long before the
values
[Demo: value iteration
k=0
Noise = 0.2
Discount = 0.9
Living reward = 0
k=1
Noise = 0.2
Discount = 0.9
Living reward = 0
k=2
Noise = 0.2
Discount = 0.9
Living reward = 0
k=3
Noise = 0.2
Discount = 0.9
Living reward = 0
k=4
Noise = 0.2
Discount = 0.9
Living reward = 0
k=5
Noise = 0.2
Discount = 0.9
Living reward = 0
k=6
Noise = 0.2
Discount = 0.9
Living reward = 0
k=7
Noise = 0.2
Discount = 0.9
Living reward = 0
k=8
Noise = 0.2
Discount = 0.9
Living reward = 0
k=9
Noise = 0.2
Discount = 0.9
Living reward = 0
k=10
Noise = 0.2
Discount = 0.9
Living reward = 0
k=11
Noise = 0.2
Discount = 0.9
Living reward = 0
k=12
Noise = 0.2
Discount = 0.9
Living reward = 0
k=100
Noise = 0.2
Discount = 0.9
Living reward = 0
Policy Iteration
Alternative approach for optimal values:
Step 1: Policy evaluation: calculate utilities for some fixed policy
(not optimal utilities!) until convergence
Step 2: Policy improvement: update policy using one-step lookahead with resulting converged (but not optimal!) utilities as future
values
Repeat steps until policy converges
This is policy iteration

Its still optimal!
Can converge (much) faster under some conditions
Policy Iteration
Evaluation: For fixed current policy , find values with policy
evaluation:
Iterate until values converge:
Improvement: For fixed values, get a better policy using policy

extraction
One-step look-ahead:
Comparison
Both value iteration and policy iteration compute the same thing (all
optimal values)
In value iteration:
Every iteration updates both the values and (implicitly) the policy
We dont track the policy, but taking the max over actions implicitly
recomputes it
In policy iteration:
We do several passes that update utilities with fixed policy (each pass is
fast because we consider only one action, not all of them)
After the policy is evaluated, a new policy is chosen (slow like a value
iteration pass)
The new policy will be better (or were done)
Summary: MDP Algorithms

So you want to.
Compute optimal values: use value iteration or policy iteration
Compute values for a particular policy: use policy evaluation
Turn your values into a policy: use policy extraction (one-step
lookahead)
These all look the same!

They basically are they are all variations of Bellman updates
They all use one-step lookahead expectimax fragments
They differ only in whether we plug in a fixed policy or max over
actions
Double Bandits
Double-Bandit MDP
Actions: Blue, Red
States: Win, Lose
0.25 $0
0.75
$2
0.25
$0
$1
1.
0
No discount
100 time
steps
Both states
have the
same value
0.75 $2
L
$1
1.
0
Offline Planning
No discount
100 time
steps
Both states
have the
same value
Solving MDPs is offline planning

You determine all quantities through computation
You need to know the details of the MDP
You do not actually play the game!
0.25 $0
Value
Play Red
Play Blue
150
100
$1
1.0
0.75
$2
0.25
$0
0.75 $2
$1
1.0
Lets Play!
$
2
$
2
$
2
$
2
$
0
$
0
$
2
$
0
$
2
$
0
Online Planning
Rules changed! Reds win chance is different.
?? $0
??
$2
??
$0
$1
1.
0
?? $2
L
$1
1.
0
Lets Play!
$
0
$
2
$
0
$
0
$
0
$
0
$
2
$
0
$
0
$
0
What Just Happened?

That wasnt planning, it was learning!
Specifically, reinforcement learning
There was an MDP, but you couldnt solve it with just computation
You needed to actually act to figure it out
Important ideas in reinforcement learning that came up
Exploration: you have to try unknown actions to get information

Exploitation: eventually, you have to use what you know
Regret: even if you learn intelligently, you make mistakes
Sampling: because of chance, you have to try things repeatedly
Difficulty: learning can be much harder than solving a known MDP
Next Time: Reinforcement Learning!

Sp14 Cs188 Lecture 9 - Mdps II

Uploaded by

Copyright:

Available Formats

Sp14 Cs188 Lecture 9 - Mdps II

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sp14 Cs188 Lecture 9 - Mdps II

Uploaded by

Copyright:

Available Formats

CS 188: Artificial Intelligence

Markov Decision Processes II

Example: Grid World

Policy = map of states to actions

The value (utility) of a q-state

[Demo: gridworld values (L9D1)]

The Bellman Equations

The Bellman Equations

Definition of optimal utility via expectimax

These are the Bellman equations, and they

Value iteration computes them:

Value iteration is just a fixed point solution method

though the Vk vectors are also interpretable as time-limited values

Utilities for a Fixed Policy

Recursive relation (one-step look-ahead / Bellman

Example: Policy Evaluation

Example: Policy Evaluation

How do we calculate the Vs for a fixed policy ?

Idea 1: Turn recursive Bellman equations into updates

Efficiency: O(S2) per iteration

Computing Actions from Values

We need to do a mini-expectimax (one step)

This is called policy extraction, since it gets the policy

Computing Actions from Q-Values

Important lesson: actions are easier to select from q-values

Problems with Value Iteration

Problem 1: Its slow O(S2A) per iteration

Problem 2: The max at each state rarely changes

This is policy iteration

Improvement: For fixed values, get a better policy using policy

Summary: MDP Algorithms

These all look the same!

Solving MDPs is offline planning

What Just Happened?

Important ideas in reinforcement learning that came up

Exploration: you have to try unknown actions to get information

Next Time: Reinforcement Learning!

You might also like