CS344M Autonomous Multiagent Systems Todd Hester Department of - - PowerPoint PPT Presentation

cs344m autonomous multiagent systems
SMART_READER_LITE
LIVE PREVIEW

CS344M Autonomous Multiagent Systems Todd Hester Department of - - PowerPoint PPT Presentation

CS344M Autonomous Multiagent Systems Todd Hester Department of Computer Science The University of Texas at Austin Good Afternoon, Colleagues Are there any questions? Todd Hester Logistics Next weeks readings up Todd Hester Logistics


slide-1
SLIDE 1

CS344M Autonomous Multiagent Systems

Todd Hester Department of Computer Science The University of Texas at Austin

slide-2
SLIDE 2

Good Afternoon, Colleagues

Are there any questions?

Todd Hester

slide-3
SLIDE 3

Logistics

  • Next week’s readings up

Todd Hester

slide-4
SLIDE 4

Logistics

  • Next week’s readings up
  • Progress reports due in 1 week

Todd Hester

slide-5
SLIDE 5

T-tests

  • Student’s T-test history

Todd Hester

slide-6
SLIDE 6

T-tests

  • Student’s T-test history
  • Paired T-test

Todd Hester

slide-7
SLIDE 7

T-tests

  • Student’s T-test history
  • Paired T-test
  • Pair tests from both distributions

Todd Hester

slide-8
SLIDE 8

T-tests

  • Student’s T-test history
  • Paired T-test
  • Pair tests from both distributions
  • Some reason to pair them... not just random

Todd Hester

slide-9
SLIDE 9

T-tests

  • Student’s T-test history
  • Paired T-test
  • Pair tests from both distributions
  • Some reason to pair them... not just random
  • Individuals before and after (pair indviduals)

Todd Hester

slide-10
SLIDE 10

T-tests

  • Student’s T-test history
  • Paired T-test
  • Pair tests from both distributions
  • Some reason to pair them... not just random
  • Individuals before and after (pair indviduals)
  • Playing

different

  • pponents

(pair games vs same

  • pponent)

Todd Hester

slide-11
SLIDE 11

T-tests

  • Student’s T-test history
  • Paired T-test
  • Pair tests from both distributions
  • Some reason to pair them... not just random
  • Individuals before and after (pair indviduals)
  • Playing

different

  • pponents

(pair games vs same

  • pponent)
  • P-value: Probability of this test result given null hypothesis

is true

Todd Hester

slide-12
SLIDE 12

T-tests

  • Student’s T-test history
  • Paired T-test
  • Pair tests from both distributions
  • Some reason to pair them... not just random
  • Individuals before and after (pair indviduals)
  • Playing

different

  • pponents

(pair games vs same

  • pponent)
  • P-value: Probability of this test result given null hypothesis

is true

Todd Hester

slide-13
SLIDE 13

T-test vs. Paired T-test

  • Is the right half of the class or the left half taller?

Todd Hester

slide-14
SLIDE 14

T-test vs. Paired T-test

  • Is the right half of the class or the left half taller?
  • Did you weigh less after the class than before?

Todd Hester

slide-15
SLIDE 15

T-test vs. Paired T-test

  • Is the right half of the class or the left half taller?
  • Did you weigh less after the class than before?
  • Who’s better at tetris?

Todd Hester

slide-16
SLIDE 16

T-test vs. Paired T-test

  • Is the right half of the class or the left half taller?
  • Did you weigh less after the class than before?
  • Who’s better at tetris?
  • Who’s better at video games in general?

Todd Hester

slide-17
SLIDE 17

T-test vs. Paired T-test

  • Test: Your team better than UvA?

Todd Hester

slide-18
SLIDE 18

T-test vs. Paired T-test

  • Test: Your team better than UvA?
  • Test:

Your team better than UvA when playing Brainstormers?

Todd Hester

slide-19
SLIDE 19

T-test vs. Paired T-test

  • Test: Your team better than UvA?
  • Test:

Your team better than UvA when playing Brainstormers?

  • Test: Your team better than UvA vs. a set of 20 opponents?

Todd Hester

slide-20
SLIDE 20

T-test vs. Paired T-test

  • Test: Your team better than UvA?
  • Test:

Your team better than UvA when playing Brainstormers?

  • Test: Your team better than UvA vs. a set of 20 opponents?
  • What if neither is significant?

Todd Hester

slide-21
SLIDE 21

T-test vs. Paired T-test

  • Test: Your team better than UvA?
  • Test:

Your team better than UvA when playing Brainstormers?

  • Test: Your team better than UvA vs. a set of 20 opponents?
  • What if neither is significant?
  • Can you run multiple times until you get a significant

result?

Todd Hester

slide-22
SLIDE 22

Other testing concepts

  • When use a chi-squared test?

Todd Hester

slide-23
SLIDE 23

Other testing concepts

  • When use a chi-squared test?
  • Discrete data classes (no such thing as a mean)

Todd Hester

slide-24
SLIDE 24

Other testing concepts

  • When use a chi-squared test?
  • Discrete data classes (no such thing as a mean)
  • Is the number of wins/ties/losses different vs. UvA than in

self play?

Todd Hester

slide-25
SLIDE 25

Examples from your projects

  • Student’s t-test?

Todd Hester

slide-26
SLIDE 26

Examples from your projects

  • Student’s t-test?
  • Paired t-test?

Todd Hester

slide-27
SLIDE 27

Examples from your projects

  • Student’s t-test?
  • Paired t-test?
  • Chi-squared test?

Todd Hester

slide-28
SLIDE 28

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

Todd Hester

slide-29
SLIDE 29

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

  • What if player 2 picks action 1 1/4 of the time?

Todd Hester

slide-30
SLIDE 30

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

  • What if player 2 picks action 1 1/4 of the time? 2.25, 2.25

Todd Hester

slide-31
SLIDE 31

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

  • What if player 2 picks action 1 1/4 of the time? 2.25, 2.25
  • Player 1 must be indifferent between actions 1 and 2

Todd Hester

slide-32
SLIDE 32

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

  • What if player 2 picks action 1 1/4 of the time? 2.25, 2.25
  • Player 1 must be indifferent between actions 1 and 2

q = 1/4

Todd Hester

slide-33
SLIDE 33

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

  • What if player 2 picks action 1 1/4 of the time? 2.25, 2.25
  • Player 1 must be indifferent between actions 1 and 2

q = 1/4

  • Player 2 must be indifferent between actions 1 and 2

Todd Hester

slide-34
SLIDE 34

Mixed strategy equilibrium

P❧❛②❡r ✷ ❆❝t✐♦♥ ✶ ❆❝t✐♦♥ ✷ ❆❝t✐♦♥ ✶ ✸✱✼ ✷✱✷ P❧❛②❡r ✶ ❆❝t✐♦♥ ✷ ✻✱✺ ✶✱✼

  • What if player 2 picks action 1 1/4 of the time? 2.25, 2.25
  • Player 1 must be indifferent between actions 1 and 2

q = 1/4

  • Player 2 must be indifferent between actions 1 and 2

p = 2/7

Todd Hester

slide-35
SLIDE 35

Correlated Equilibria

Todd Hester

slide-36
SLIDE 36

Correlated Equilibria

Sometimes mixing isn’t enough: Bach/Stravinsky

❲✐❢❡ ❙ ❇ ❙ ✷✱✶ ✵✱✵ ▼❡ ❇ ✵✱✵ ✶✱✷

Want only S,S or B,B - 50% each

Todd Hester

slide-37
SLIDE 37

Focal points

  • We will both be in Paris for some time in June.
  • We both know that we will both be there on the 15th.

Todd Hester

slide-38
SLIDE 38

Focal points

  • We will both be in Paris for some time in June.
  • We both know that we will both be there on the 15th.
  • Something happens so that we must meet on that day
  • We have no way of getting in touch.

Todd Hester

slide-39
SLIDE 39

Focal points

  • We will both be in Paris for some time in June.
  • We both know that we will both be there on the 15th.
  • Something happens so that we must meet on that day
  • We have no way of getting in touch.
  • When and where?

Todd Hester

slide-40
SLIDE 40

Focal points

  • We will both be in Paris for some time in June.
  • We both know that we will both be there on the 15th.
  • Something happens so that we must meet on that day
  • We have no way of getting in touch.
  • When and where?
  • What are the Nash equilibria?

Todd Hester

slide-41
SLIDE 41

Focal points

  • We will both be in Paris for some time in June.
  • We both know that we will both be there on the 15th.
  • Something happens so that we must meet on that day
  • We have no way of getting in touch.
  • When and where?
  • What are the Nash equilibria?
  • Where would you meet in Austin?

Todd Hester

slide-42
SLIDE 42

Focal points

  • We will both be in Paris for some time in June.
  • We both know that we will both be there on the 15th.
  • Something happens so that we must meet on that day
  • We have no way of getting in touch.
  • When and where?
  • What are the Nash equilibria?
  • Where would you meet in Austin? New York City?

Todd Hester

slide-43
SLIDE 43

Incomplete Information Games

  • We each get one of 3 cards: 1,2,3
  • If we both fold, we both lose nothing
  • If one raises and one folds, the raiser gets 1
  • If both raise, the one with the higher card gets 5
  • Zero sum

Todd Hester

slide-44
SLIDE 44

Incomplete Information Games

  • We each get one of 3 cards: 1,2,3
  • If we both fold, we both lose nothing
  • If one raises and one folds, the raiser gets 1
  • If both raise, the one with the higher card gets 5
  • Zero sum

❈❛r❞ ❄ ❘ ❋ ❘ ✺✱✲✺ ✶✱✲✶ ❈❛r❞ ✸ ❋ ✲✶✱✶ ✵✱✵

Todd Hester

slide-45
SLIDE 45

Incomplete Information Games

❈❛r❞ ❄ ❘ ❋ ❘ ✺✱✲✺ ✶✱✲✶ ❈❛r❞ ✸ ❋ ✲✶✱✶ ✵✱✵

Todd Hester

slide-46
SLIDE 46

Incomplete Information Games

❈❛r❞ ❄ ❘ ❋ ❘ ✺✱✲✺ ✶✱✲✶ ❈❛r❞ ✸ ❋ ✲✶✱✶ ✵✱✵ ❈❛r❞ ❄ ❘ ❋ ❘ ✲✺✱✺ ✶✱✲✶ ❈❛r❞ ✶ ❋ ✲✶✱✶ ✵✱✵

Todd Hester

slide-47
SLIDE 47

Bayes-Nash Equilibrium

  • 3 ⇒ raise

Todd Hester

slide-48
SLIDE 48

Bayes-Nash Equilibrium

  • 3 ⇒ raise
  • 1 ⇒ fold (no matter what the other one does with 2)

Todd Hester

slide-49
SLIDE 49

Bayes-Nash Equilibrium

  • 3 ⇒ raise
  • 1 ⇒ fold (no matter what the other one does with 2)
  • 2 ⇒ ?

Todd Hester

slide-50
SLIDE 50

Bayes-Nash Equilibrium

  • 3 ⇒ raise
  • 1 ⇒ fold (no matter what the other one does with 2)
  • 2 ⇒ ?

− Raise: (.5)(-5) + (.5)(1) = -2 − Fold: (.5)(-1) + (.5)(0) = -.5

Todd Hester

slide-51
SLIDE 51

Bayes-Nash Equilibrium

  • 3 ⇒ raise
  • 1 ⇒ fold (no matter what the other one does with 2)
  • 2 ⇒ ?

− Raise: (.5)(-5) + (.5)(1) = -2 − Fold: (.5)(-1) + (.5)(0) = -.5 − Always fold!

Todd Hester

slide-52
SLIDE 52

Bayes-Nash Equilibrium

  • 3 ⇒ raise
  • 1 ⇒ fold (no matter what the other one does with 2)
  • 2 ⇒ ?

− Raise: (.5)(-5) + (.5)(1) = -2 − Fold: (.5)(-1) + (.5)(0) = -.5 − Always fold! − Bayes-Nash: both players Raise if 3, otherwise Fold

Todd Hester

slide-53
SLIDE 53

Bayes-Nash Equilibrium

  • 3 ⇒ raise
  • 1 ⇒ fold (no matter what the other one does with 2)
  • 2 ⇒ ?

− Raise: (.5)(-5) + (.5)(1) = -2 − Fold: (.5)(-1) + (.5)(0) = -.5 − Always fold! − Bayes-Nash: both players Raise if 3, otherwise Fold With more numbers and/or different payoffs, bluffing can be a part of the Nash Equilibrium

Todd Hester

slide-54
SLIDE 54

Discussion

  • How useful is the concept of Nash equilibrium?

Todd Hester

slide-55
SLIDE 55

Discussion

  • How useful is the concept of Nash equilibrium?
  • What if one player isn’t rational?

Todd Hester

slide-56
SLIDE 56

Discussion

  • How useful is the concept of Nash equilibrium?
  • What if one player isn’t rational?
  • What can’t game theory simulate?

Todd Hester