Evolving New Strategies - Computer Science & Engineering
Download
Report
Transcript Evolving New Strategies - Computer Science & Engineering
Evolving New Strategies
The Evolution of Strategies in the
Iterated Prisoner’s Dilemma
01 / 25
What is the Prisoner’s Dilemma?
There are two prisoners
Each one has taken part in the same criminal act
The authorities are interrogating each one
Each prisoner can choose to keep their mouth shut or rat out their
partner
If both prisoners stay quiet, they each get n months of jail time
If only one prisoner gets ratted out, that prisoner gets n + x months
of jail time while the other prisoner gets n – y months of jail time
If the prisoners rat each other out, they each get n + z months of jail
time.
In this case, n, x, y, and z are all greater than zero.
In this case, x is greater than z.
02 / 25
What is the Iterated Prisoner’s
Dilemma?
Prisoner’s Dilemma performed several times
The two criminals have committed several
crimes together
They are interrogated for each crime, with each
set of interrogations being an instance of the
original Prisoner’s Dilemma
These interrogations are performed in sequence
(or iteratively), and the jail time distributed to
each prisoner is cumulative
03 / 25
How does the IPD relate to GAs?
No optimal solution
No real strategy
No clue
Hard problem
So back to the paper
04 / 25
What This Paper Shows
GAs in a rich social setting
Advantage of developing new strategies
One parent
Two parent
Early commitments to paths
Evolutionary processes optimal or arbitrary
05 / 25
How Does It Show It?
Simulation
Multiple cases
Comparative output
06 / 25
The Simulation
Specify the environment
Specify the encoding
Testing the effects of random mutation
Run the simulation
Analyze the results
07 / 25
The Environment
Prisoner’s dilemma
Multiple prisoners
Goal is to achieve mutual cooperation
Individuals may meet more than once
08 / 25
Initial Experiment
Original strategies were submitted by
fourteen people
Game Theory
Economics
Sociology
Political Science
Mathematics
Various levels of intricacy
09 / 25
Initial Experiment
Most complex strategy
Markov process
Bayesian inference
Least complex strategy
TFT
TFT won
10 / 25
Second Experiment
Sixty-two entries
Six countries
Computer hobbyists, professors
TFT was submitted again
It won
11 / 25
The GA
Population
Encoding
Generation
Crossover
Mutation
Fifty Generations
12 / 25
Population
Twenty chromosomes
Seventy genes
13 / 25
Encoding
For each prisoner’s dilemma, there are
four possibilities
Each “player” has memory
What each gene represents
14 / 25
A Single Generation
Multiple games
Each game had one-hundred and fifty-one moves
Each chromosome played eight others
Fitness was assigned
Ratted out – Zero points
Mutually ratted out – One point
Mutual cooperation – Three points
You ratted, other person stayed quit – Five points
15 / 25
Crossover
Fitness proportional selection
Involved standard deviation from mean
Strictly ten crossovers
Single point
Two parents
16 / 25
Mutation
Single gene flip
One gene per two chromosomes
17 / 25
Results
Median resultant member
Just as good as TFT
Resembled TFT
Five properties were found
Don’t rock the boat
Be provocable
Accept apologies
Forget
Accept a rut
18 / 25
Results
ADJUSTER
Special chromosome which consistently
seeks to exploit
TFT
Majority of other chromosomes
19 / 25
Results
Twenty-five percent of runs
Median was better
Exploit one chromosome
20 / 25
Results: Why is this important?
Chromosomes had to learn
Discriminatory based on evidence
Self adjusting for exploitation
No alienation
Break primary rule of first tournament
Be nice? I don’t think so
21 / 25
Results: Misleading?
Median
Exploitative
Menacing
A true criminal?
Fixed size population and tournaments
Simulate real evolution
22 / 25
Results: A Slight Twist
Asexual reproduction
TFT
Less than half of the medians
Changing environment
Play against everyone
Everyone starts aggressive
Fitness
rapidly declines
Fitness begins to even out
Fitness begins to rise
23 / 25
Conclusions
The GA is good for searching, large, multidimensional spaces
Multiple parent crossover helps
Arbitrary aspects of evolution
Exploration vs. Exploitation
Hitch hikers
Selection Pressure
Evolutionary Commitments can be
irreversible
24 / 25
Related Topics
Mutation
Crossover
Inversion
Coding principles
Dominant/Recessive
Rate of evolution
Population viscosity
Speciation and niches
25 / 25