Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Evolutionary Game Theory
1. DATA MINING AND MACHINE LEARNING
IN A NUTSHELL
EVOLUTIONARY GAME THEORY
Mohammad-Ali Abbasi
http://www.public.asu.edu/~mabbasi2/
SCHOOL OF COMPUTING, INFORMATICS, AND DECISION SYSTEMS ENGINEERING
ARIZONA STATE UNIVERSITY
Arizona State University
http://dmml.asu.edu/
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 1
2. What is a Game
• Game theory is concerned with situations in
which decision-makers interact with one
another,
• and in which the happiness of each participant
with the outcome depends not just on his or
her own decisions but on the decisions made
by everyone.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 2 2
3. What is game theory?
• Formal way to analyse interactions between
agents who behave strategically
• Mathematics of decision making in conflict
situations
• Usual to assume players are “rational”
• Widely applied to the study of economics,
warfare, politics, animal behaviour,
sociology, business, ecology and
evolutionary biology
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 3
4. Basic elements of a Game
• Players
– Everyone who has an effect on your earnings
• Strategies
– Actions available to each player
– Define a plan of action for every contingency
• Payoffs
– Numbers associated with each outcome
– Reflect the interests of the players
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 4
5. Nash Equilibrium
• A Nash equilibrium is a situation in which
none of them have dominant Strategy and
each player makes his or her best response
– (S, T) is Nash equilibrium if S is the best strategy to
T and T is the best strategy to S
• John Nash shared the 1994 Nobel prize in
Economic for developing this idea!
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 5
6. Prisoner’s Dilemma
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 6
7. Evolutionary Game Theory
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 7
8. Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 8
9. evolutionary stable strategy
• Taller trees get more light, so taller trees
reproduce more.
• Taller trees have to consume more resources
to be tall.
• System converges to a state where only the
tallest trees are present.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 9
10. Evolutionary stable strategy
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 10
11. Evolutionary Game Theory
1 x1 x1
1 1/ r r…relative fitness of new mutant
n n…population size
1 1/ r
For a neutral mutant, r=1, the fixation probability is 1/n.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 11
12. Evolutionary Game Theory
• Is the attempt to invent and study
mathematical equations describing
how population change over time due
to mutation and selection (Learning).
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 12
13. GT vs. EGT
• In GT, one assumes that agents are perfectly
rational.
• In EGT, trial and error process gives strategies
that can be selected for by some force
(evolution - biological, cultural, etc…).
• This lack of rationality is the point of
departure between EGT and GT.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 13
14. Evolutionary game theory
• population of players
• follow different strategies
• frequencies of strategies change over time,
dependent on success relative to other
strategies
• genetic inheritance (mutation) or learning
(innovation)
– Irrational
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 14
15. Evolutionary Biology
Evolutionary biology is based on the idea that an
organism's genes largely determine its observable
characteristics, and hence its fitness in a given
environment.
Organisms that are more fit will tend to produce
more offspring, causing genes that provide
greater fitness to increase their representation in
the population.
In this way, fitter genes tend to win over time,
because they provide higher rates of
reproduction.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 15
16. Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 16
17. Competition for food
• When beetles of the same size compete, they
get equal shares of the food
• When a large beetle competes with a small
beetle, the large beetle gets the majority of
the food.
• In all cases, large beetles experience less of a
fitness benefit from a given quantity of food,
since some of it is diverted into maintaining
their expensive metabolism
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 17
18. Body Size Game
Beetle 2
Small Large
Small 5, 5 1, 8
Beetle 1
Large 8, 1 3, 3
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 18
19. • Small portion of Large Beetles: x
• Small Beetles: 1-x
• Expected Pay off in a population that small is
majority
– Small beetle
• 5(1-x) + 1.X = 5- 4x
– Large beetle
• 8 (1- x) + 3.x = 8-5x
• Small is not evolutionary stable!
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 19
20. • Expected Pay off in a population that Large is
majority
• Large Beetle: 3 * (1-x) + 8 * x = 3 + 5x
• Small Beetle: 1 * (1-x) + 5 * x = 1 + 4x
• Large is evolutionary stable
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 20
21. Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 21
22. Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 22
23. Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 23
24. General Description
Organism 2
S T
S a, a b, c
Organism 1
T c, b d, d
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 24
25. • X -> T
• (1-x) -> S
• Play S
– Expected Payoff: a (1-x) + bx
• Play T
– Expected Payoff: C ( 1- x) + dx
A ( 1- x) + bx > c ( 1-x ) + dx
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 25
26. • In a two-player, two-strategy, symmetric
game, S is evolutionarily stable precisely when
either
a>c
or
a = c and b > d.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 26
27. Relationship between evolutionary and Nash
Equilibria
• (S, S) is a Nash equilibrium when S is a best
response to the choice of S by the other player
a >= c
• The condition for S to be evolutionarily stable
a>c
Or
a = c and b > d
• If strategy S is evolutionarily stable, then (S, S) is a
Nash equilibrium
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 27
28. • Other direction
– (S, S) is a Nash equilibrium -> S is not ESS
a = c and b < d
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 28
29. Strict Nash Equilibrium
Hunt Hunter 2
Stag or Hare S H
S 4, 4 0, 3
Hunter 1
H 3, 0 3, 3
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 29
30. Strict Nash Equilibrium
Hunt Hunter 2
Stag or Hare S H
S 4, 4 0, 4
Hunter 1
H 4, 0 3, 3
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 30
31. Strict Nash Equilibrium
choice of strategies is a strict Nash equilibrium if
each player is using the unique best response to
what the other player is doing
for symmetric two-player, two-strategy games, the
condition for (S, S) to be a strict Nash equilibrium
is that a > c
the set of evolutionarily stable strategies S is a
subset of the set of strategies S for which (S, S) is
a Nash equilibrium
if (S, S) is a strict Nash equilibrium, then S is
evolutionarily stable
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 31
32. Nash Equilibrium and Evolutionary Stability
• In a Nash equilibrium
– we consider players choosing mutual best responses
to each other's strategy
– This equilibrium concept places great demands on the
ability of the players to chose optimally and to
coordinate on strategies that are best responses to
each other.
• Evolutionary stability
– no intelligence or coordination on the part of the
players
– strategies are viewed as being hard-wired into the
players, perhaps because their behavior is encoded in
their genes
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 32
33. Evolutionarily Stable Mixed
Strategies
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 33
34. General Description
Player 2
S T
S a, a b, c
Player 1
T c, b d, d
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 34
35. Evolutionarily Stable Mixed Strategies
Organism 1:
Play S with probability p and T with (1-p)
Organism 2:
Play S with probability q and T with (1-q)
V (p, q) = pqa + p(1-q)b + (1-p)qc + (1-p)(1-q)d
For p to be ESMS
(1-x)V(p, p) + xV(p, q) > (1-x) V(q, p) + xV(q, q)
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 35
36. Evolutionarily Stable Mixed Strategies
• In the General Symmetric Game, p is an
evolutionarily stable mixed strategy if there is
a (small) positive number y such that when
any other mixed strategy q invades p at any
level x < y, the fitness of an organism playing p
is strictly greater than the fitness of an
organism playing q.
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 36
37. Mohammad-Ali Abbasi (Ali),
Ali, is a Ph.D student at Data Mining
and Machine Learning Lab, Arizona
State University.
His research interests include Data
Mining, Machine Learning, Social
Computing, and Social Media Behavior
Analysis.
http://www.public.asu.edu/~mabbasi2/
Arizona State University
Data Mining and Machine Learning Lab
Data Mining and Machine Learning- in a nutshell Evolutionary Game Theory 37
Notas do Editor
There is a set of participants, whom we call the playersIn our example, you and yourpartner are the two players.(ii) Each player has a set of options for how to behave; we will refer to these as the player'spossible strategiesIn the example, you and your partner each have two possiblestrategies: to prepare for the presentation, or to study for the exam.(iii) For each choice of strategies, each player receives a payo that can depend on thestrategies selected by everyoneThepayos will generally be numbers, with eachplayer preferring larger payos to smaller payosIn our current example, the payoto each player is the average grade he or she gets on the exam and the presentation