Reinforcement learning presentation1.ppt

Reinforcement
Learning
Yijue Hou

What is learning?
 Learning takes place as a result of
interaction between an agent and the
world, the idea behind learning is that
 Percepts received by an agent should be used
not only for acting, but also for improving the
agent’s ability to behave optimally in the future
to achieve the goal.

Learning types
 Learning types
 Supervised learning:
a situation in which sample (input, output) pairs of the
function to be learned can be perceived or are given

You can think it as if there is a kind teacher
 Reinforcement learning:
in the case of the agent acts on its environment, it
receives some evaluation of its action
(reinforcement), but is not told of which action is the
correct one to achieve its goal

Reinforcement learning
 Task
Learn how to behave successfully to achieve a
goal while interacting with an external
environment
 Learn via experiences!
 Examples
 Game playing: player knows whether it win or
lose, but not know how to move at each step
 Control: a traffic system can measure the delay
of cars, but not know how to decrease it.

RL is learning from
interaction

RL model
 Each percept(e) is enough to determine the
State(the state is accessible)
 The agent can decompose the Reward
component from a percept.
 The agent task: to find a optimal policy,
mapping states to actions, that maximize long-
run measure of the reinforcement
 Think of reinforcement as reward
 Can be modeled as MDP model!

Review of MDP model
 MDP model <S,T,A,R>
Agent
Environment
State
Reward
Action
s0
r0
a0
s1
a1
r1
s2
a2
r2
s3
• S– set of states
• A– set of actions
• T(s,a,s’) = P(s’|s,a)– the
probability of transition from
s to s’ given action a
• R(s,a)– the expected reward
for taking action a in state s




'
'
)
'
,
,
(
)
'
,
,
(
)
,
(
)
'
,
,
(
)
,
|
'
(
)
,
(
s
s
s
a
s
r
s
a
s
T
a
s
R
s
a
s
r
a
s
s
P
a
s
R

Model based v.s.Model free
approaches
 But, we don’t know anything about the
environment model—the transition function
T(s,a,s’)
 Here comes two approaches
 Model based approach RL:
learn the model, and use it to derive the optimal policy.
e.g Adaptive dynamic learning(ADP) approach
 Model free approach RL:
derive the optimal policy without learning the model.
e.g LMS and Temporal difference approach
 Which one is better?

Passive learning v.s. Active
learning
 Passive learning
 The agent imply watches the world going by
and tries to learn the utilities of being in
various states
 Active learning
 The agent not simply watches, but also acts

Passive learning scenario
 The agent see the the sequences of state trans
itions and associate rewards
 The environment generates state transitions and the
agent perceive them
e.g (1,1) (1,2) (1,3) (2,3) (3,3) (4,3)[+1]
(1,1)(1,2) (1,3) (1,2) (1,3) (1,2) (1,1) (2,1) (3,
1) (4,1) (4,2)[-1]
 Key idea: updating the utility value using the gi
ven training sequences.

LMS updating
 Reward to go of a state
the sum of the rewards from that state until a
terminal state is reached
 Key: use observed reward to go of the state as
the direct evidence of the actual expected
utility of that state
 Learning utility function directly from
sequence example

LMS updating
function LMS-UPDATE (U, e, percepts, M, N ) return an updated U
if TERMINAL?[e] then
{ reward-to-go  0
for each ei in percepts (starting from end) do
s = STATE[ei]
reward-to-go  reward-to-go + REWARS[ei]
U[s] = RUNNING-AVERAGE (U[s], reward-to-go, N[s])
end
}
function RUNNING-AVERAGE (U[s], reward-to-go, N[s] )
U[s] = [ U[s] * (N[s] – 1) + reward-to-go ] / N[s]

LMS updating algorithm in
passive learning
 Drawback:
 The actual utility of a state is constrained to be probability- weighted
average of its successor’s utilities.
 Converge very slowly to correct utilities values (requires a lot of
sequences)

for our example, >1000!

Temporal difference method
in passive learning
 TD(0) key idea:
 adjust the estimated utility value of the current state based on its im
mediately reward and the estimated value of the next state.
 The updating rule
 is the learning rate parameter
 Only when is a function that decreases as the number of times
a state has been visited increased, then can U(s)converge to the c
orrect value.
))
(
)
'
(
)
(
(
)
(
)
( s
U
s
U
s
R
s
U
s
U 


 



The TD learning curve
(4,3)
(2,3)
(2,2)
(1,1)
(3,1)
(4,1)
(4,2)

Adaptive dynamic
programming(ADP) in passive
learning
 Different with LMS and TD method(model free
approaches)
 ADP is a model based approach!
 The updating rule for passive learning
 However, in an unknown environment, T is not
given, the agent must learn T itself by
experiences with the environment.
 How to learn T?
))
'
(
)
'
,
(
(
)
'
,
(
)
(
'
s
U
s
s
r
s
s
T
s
U
s




ADP learning curves
(4,3)
(3,3)
(2,3)
(1,1)
(3,1)
(4,1)
(4,2)

Active learning
 An active agent must consider
 what actions to take?
 what their outcomes maybe(both on learning and receiving the
rewards in the long run)?
 Update utility equation
 Rule to chose action
))
'
(
)
'
,
,
(
)
,
(
(
max
arg
'
s
U
s
a
s
T
a
s
R
a
s
a


 
))
'
(
)
'
,
,
(
)
,
(
(
max
)
(
'
s
U
s
a
s
T
a
s
R
s
U
s
a


 

Active ADP algorithm
For each s, initialize U(s) , T(s,a,s’) and R(s,a)
Initialize s to current state that is perceived
Loop forever
{
Select an action a and execute it (using current model R and T) using
Receive immediate reward r and observe the new state s’
Using the transition tuple <s,a,s’,r> to update model R and T (see further)
For all the sate s, update U(s) using the updating rule
s = s’
}
))
'
(
)
'
,
,
(
)
,
(
(
max
arg
'
s
U
s
a
s
T
a
s
R
a
s
a


 
))
'
(
)
'
,
,
(
)
,
(
(
max
)
(
'
s
U
s
a
s
T
a
s
R
s
U
s
a


 

How to learn model?
 Use the transition tuple <s, a, s’, r> to learn T(s,a,s’) and
R(s,a). That’s supervised learning!
 Since the agent can get every transition (s, a, s’,r) directly, so
take (s,a)/s’ as an input/output example of the transition
probability function T.
 Different techniques in the supervised learning(see further
reading for detail)
 Use r and T(s,a,s’) to learn R(s,a)


'
)
'
,
,
(
)
,
(
s
r
s
a
s
T
a
s
R

ADP approach pros and cons
 Pros:
 ADP algorithm converges far faster than LMS and Temporal
learning. That is because it use the information from the the
model of the environment.
 Cons:
 Intractable for large state space
 In each step, update U for all states
 Improve this by prioritized-sweeping (see further reading for
detail)

Another model free
method– TD-Q learning
 Define Q-value function
 Q-value function updating rule
<*>
 Key idea of TD-Q learning

Combined with temporal difference approach

The updating rule
 Rule to chose the action to take
)
,
(
max
)
( a
s
Q
s
U
a

))
'
(
)
'
,
,
(
)
,
(
(
max
)
(
'
s
U
s
a
s
T
a
s
R
s
U
s
a


 
)
'
(
)
'
,
,
(
)
,
(
)
,
(
'
s
U
s
a
s
T
a
s
R
a
s
Q
s


 
)
'
,
'
(
max
)
'
,
,
(
)
,
(
)
,
(
'
'
a
s
Q
s
a
s
T
a
s
R
a
s
Q
s
a


 
))
,
(
)
'
,
'
(
max
(
)
,
(
)
,
(
'
a
s
Q
a
s
Q
r
a
s
Q
a
s
Q
a



 

)
,
(
max
arg a
s
Q
a
a


TD-Q learning agent
algorithm
For each pair (s, a), initialize Q(s,a)
Observe the current state s
Loop forever
{
Select an action a and execute it
Receive immediate reward r and observe the new state s’
Update Q(s,a)
s=s’
}
)
,
(
max
arg a
s
Q
a
a

))
,
(
)
'
,
'
(
max
(
)
,
(
)
,
(
'
a
s
Q
a
s
Q
r
a
s
Q
a
s
Q
a



 


 An action has two kinds of outcome
 Gain rewards on the current
experience tuple (s,a,s’)
 Affect the percepts received, and
hence the ability of the agent to learn
Exploration problem in
Active learning

Active learning
 A trade off when choosing action between
 its immediately good(reflected in its current utility estimates using
the what we have learned)
 its long term good(exploring more about the environment help it to
behave optimally in the long run)
 Two extreme approaches
 “wacky”approach: acts randomly, in the hope that it will eventually
explore the entire environment.
 “greedy”approach: acts to maximize its utility using current model
estimate
See Figure 20.10
 Just like human in the real world! People need to decide between
 Continuing in a comfortable existence
 Or striking out into the unknown in the hopes of discovering a new
and better life

Active learning
 One kind of solution: the agent should be more wacky when it
has little idea of the environment, and more greedy when it
has a model that is close to being correct
 In a given state, the agent should give some weight to actions
that it has not tried very often.
 While tend to avoid actions that are believed to be of low utility
 Implemented by exploration function f(u,n):
 assigning a higher utility estimate to relatively unexplored action
state pairs
 Chang the updating rule of value function to
 U+ denote the optimistic estimate of the utility
))
,
(
),
'
(
)
'
,
,
(
(
)
,
(
(
max
)
(
'
s
a
N
s
U
s
a
s
T
f
a
s
r
s
U
s
a




 

Active learning
 One kind of definition of f(u,n)
if n< Ne
u otherwise
 is an optimistic estimate of the best possible reward obtainable
in any state
 The agent will try each action-state pair(s,a) at least Ne times
 The agent will behave initially as if there were wonderful rewards sc
attered all over around– optimistic .

)
,
( n
u
f

R


R

Generalization in
Reinforcement Learning
 So far we assumed that all the functions learned
by the agent are (U, T, R,Q) are tabular forms—
i.e.. It is possible to enumerate state and action
spaces.
 Use generalization techniques to deal with large
state or action space.
 Function approximation techniques

Genetic algorithm and Evolutionary
programming
 Start with a set of individuals
 Apply selection and reproduction operators to “evolve” an individual that is
successful (measured by a fitness function)

programming
 Imagine the individuals as agent functions
 Fitness function as performance measure or
reward function
 No attempt made to learn the relationship the
rewards and actions taken by an agent
 Simply searches directly in the individual space to
find one that maximizes the fitness functions

programming
 Represent an individual as a binary string(each bit of the string is called a
gene)
 Selection works like this: if individual X scores twice as high as Y on the
fitness function, then X is twice likely to be selected for reproduction than Y
is
 Reproduction is accomplished by cross-over and mutation

Reinforcement learning presentation1.ppt

More Related Content

Similar to Reinforcement learning presentation1.ppt (20)

Recently uploaded (20)

Reinforcement learning presentation1.ppt