Classifier Classifier Systems Systems - - PowerPoint PPT Presentation

▶
classifier classifier systems systems
SMART_READER_LITE
LIVE PREVIEW

Classifier Classifier Systems Systems - - PowerPoint PPT Presentation

Nov 18, 2023 9 likes •291 views

Classifier Classifier Systems Systems Christian Jacob Christian Jacob jacob@cpsc.ucalgary.ca Department of Computer Science University of Calgary

slide-1
SLIDE 1

Classifier Classifier Systems Systems

—————————————— —————————————— Christian Jacob Christian Jacob

jacob@cpsc.ucalgary.ca

Department of Computer Science University of Calgary

slide-2
SLIDE 2

Cellular Automata Random Boolean Networks Classifier Systems Swarm Systems

2

slide-3
SLIDE 3

Classifier Classifier Systems Systems

  • J. Holland (1975)
  • J. Holland (1975)

Learning syntactically Learning syntactically simple simple string rules string rules ( (classifiers classifiers) to ) to guide guide performance performance in an in an arbitrary arbitrary environment environment

slide-4
SLIDE 4

Objective: A Formal Framework for an Objective: A Formal Framework for an Operon Operon-Operator Gene Regulation Model

  • Operator Gene Regulation Model

( (Britten Britten-Davidson)

  • Davidson)
  • J. Holland: Adaptation in Natural and Artificial Systems

4

slide-5
SLIDE 5

First a Simple Example ... First a Simple Example ...

F F A classifier system to emulate a frog.

A classifier system to emulate a frog. The frog reacts to objects it sees. The frog reacts to objects it sees.

Moving On the Ground Large Far Striped Flee! Pursue!

Input: Output:

1 _ _ _ _ 1 1 _ 1 1 1

slide-6
SLIDE 6

Classifier System in Action Classifier System in Action

Environ- mental Signal Action

Detectors 1 _ 1 Message List Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x Effectors 1 1 _

101

slide-7
SLIDE 7

Classifier System in Action Classifier System in Action

Environ- mental Signal Action

Detectors 1 _ 1 Message List 1 0 1 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x Effectors 1 1 _

slide-8
SLIDE 8

Classifier System in Action Classifier System in Action

Environ- mental Signal Action

Message List 1 1 1 0 0 0 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x Detectors 1 _ 1 Effectors 1 1 _

111

slide-9
SLIDE 9

Classifier System in Action Classifier System in Action

Environ- mental Signal Action

Message List 0 0 0 0 0 1 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x Detectors 1 _ 1 Effectors 1 1 _

slide-10
SLIDE 10

Classifier System in Action Classifier System in Action

Environ- mental Signal Action

Message List 0 0 0 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x Detectors 1 _ 1 Effectors 1 1 _

slide-11
SLIDE 11

Classifier System in Action Classifier System in Action

Environ- mental Signal Action

Message List 0 0 0 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x Detectors 1 _ 1 Effectors 1 1 _

How can we adapt this rule set?

slide-12
SLIDE 12

Learning CS Architecture Learning CS Architecture

Environ- mental Signal Action

Detectors 1 _ 1 Effectors 1 1 _ Message List 1 0 1 0 0 0 1 1 1 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x

Genetic Algorithm 101

slide-13
SLIDE 13

Genetic Algorithms Genetic Algorithms

  • J. Holland (1975)
  • J. Holland (1975)
  • D. Goldberg (1989)
  • D. Goldberg (1989)

Simulated Genome Simulated Genome-

  • based

based Evolution Evolution

slide-14
SLIDE 14

{1,0,1,1,0,1,0,0,1,0,1,1} {0,1,1,1,1,0,0,1,0,0,0,1} {0,0,1,1,0,101,1,0,1,0,0} ... {1,1,0,0,0,1,0,1,0,1,0,0} ... {1,0,1,0,0,1,1,1,0,1,1,1} {0,0,1,1,0,1,1,1,0,1,0,0} {1,0,0,1,0,1,1,1,0,0,0,1}

Binary vector Binary vector decoding interpretation

Genetic Algorithms Genetic Algorithms

Representation of individuals Representation of individuals

slide-15
SLIDE 15

{1,0,1,1,0,1,0,0,1,0,1,1} {0,1,1,1,1,0,0,1,0,0,0,1} {1,1,0,0,0,1,0,1,0,1,0,0} ... {1,0,1,0,0,1,1,1,0,1,1,1} {0,0,1,1,0,1,1,1,0,1,0,0} {1,0,0,1,0,1,1,1,0,0,0,1}

{1,1,0,0,0,1,0,1,0,1,0,0} {0,0,1,1,0,1,1,1,0,1,0,0} {1,1,1,1,0,1,0,1,0,0,0,0} {0,1,1,1,0,0,1,1,0,1,1,0}

selection selection mutation mutation

{0,1,1,1,0,1,0,1,0,0,0,0} {1,1,1,1,0,0,1,1,0,1,1,0}

crossover crossover

{1,0,1,1,0,1,0,0,1,0,1,1} {0,1,1,1,1,0,0,1,0,0,0,1} {1,1,0,0,0,1,0,1,0,1,0,0} ... {1,0,1,0,0,1,1,1,0,1,1,1} {0,0,1,1,0,1,1,1,0,1,0,0} {1,0,0,1,0,1,1,1,0,0,0,1}

interpretation evaluation

  • Ind. 40
  • Ind. 38
  • Ind. 7
  • Ind. 5
  • Ind. 3
  • Ind. 1

2 4 6

  • Ind. 40
  • Ind. 38
  • Ind. 7
  • Ind. 5
  • Ind. 3
  • Ind. 1
slide-16
SLIDE 16

Learning CS Architecture Learning CS Architecture

Environ- mental Signal Action

Detectors 1 _ 1 Effectors 1 1 _ Message List 1 0 1 0 0 0 1 1 1 Classifiers 1 0 _ : 1 1 1 0 0 _ : 0 0 0 1 x_ 1 : 0 0 x

Genetic Algorithm 101

slide-17
SLIDE 17

How do Classifiers Receive How do Classifiers Receive their Fitnesses their Fitnesses? ?

Apportionment of Credit Apportionment of Credit through through Bucket Brigades Bucket Brigades

slide-18
SLIDE 18

Bucket Brigade Algorithm Bucket Brigade Algorithm

Index Rule Fitness Triggering Bid Message Rule _______________________________________________________ 1 0 1 _ _ : 0000 200 20 0000 2 0 0 _ 0 : 1100 200 3 1 1 _ _ : 1000 200 4 _ _ 0 0 : 0001 200 ––––––––––––––––––––––––––––––––––––––––––––––––––––––– 1 0 1 _ _ : 0000 180 2 0 0 _ 0 : 1100 200 1 20 1100 3 1 1 _ _ : 1000 200 4 _ _ 0 0 : 0001 200 1 20 0001 –––––––––––––––––––––––––––––––––––––––––––––––––––––––

1 2

slide-19
SLIDE 19

Bucket Brigade Algorithm Bucket Brigade Algorithm

Index Rule Fitness Triggering Bid Message Rule _______________________________________________________ 1 0 1 _ _ : 0000 180 2 0 0 _ 0 : 1100 200 1 20 1100 3 1 1 _ _ : 1000 200 4 _ _ 0 0 : 0001 200 1 20 0001 ––––––––––––––––––––––––––––––––––––––––––––––––––––––– 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 180 3 1 1 _ _ : 1000 200 2 20 1000 4 _ _ 0 0 : 0001 180 2 18 0001 –––––––––––––––––––––––––––––––––––––––––––––––––––––––

2 3

slide-20
SLIDE 20

Bucket Brigade Algorithm Bucket Brigade Algorithm

Index Rule Fitness Triggering Bid Message Rule _______________________________________________________ 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 180 3 1 1 _ _ : 1000 200 2 20 1000 4 _ _ 0 0 : 0001 180 2 18 0001 ––––––––––––––––––––––––––––––––––––––––––––––––––––––– 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 218 3 1 1 _ _ : 1000 180 4 _ _ 0 0 : 0001 162 3 16 0001 –––––––––––––––––––––––––––––––––––––––––––––––––––––––

3 4

slide-21
SLIDE 21

Bucket Brigade Algorithm Bucket Brigade Algorithm

Index Rule Fitness Triggering Bid Message Rule _______________________________________________________ 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 218 3 1 1 _ _ : 1000 180 4 _ _ 0 0 : 0001 162 3 16 0001 ––––––––––––––––––––––––––––––––––––––––––––––––––––––– 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 218 3 1 1 _ _ : 1000 196 4 _ _ 0 0 : 0001 146 –––––––––––––––––––––––––––––––––––––––––––––––––––––––

4 5

slide-22
SLIDE 22

Bucket Brigade Algorithm Bucket Brigade Algorithm

Index Rule Fitness Triggering Bid Message Rule _______________________________________________________ 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 218 3 1 1 _ _ : 1000 180 4 _ _ 0 0 : 0001 162 3 16 0001 ––––––––––––––––––––––––––––––––––––––––––––––––––––––– 1 0 1 _ _ : 0000 220 2 0 0 _ 0 : 1100 218 3 1 1 _ _ : 1000 196 4 _ _ 0 0 : 0001 146 –––––––––––––––––––––––––––––––––––––––––––––––––––––––

4 5 Here are the fitnesses

slide-23
SLIDE 23

The Broadcast Language The Broadcast Language

  • J. Holland (1975)
  • J. Holland (1975)

A Formal A Formal Framework for Modeling Framework for Modeling Evolvable Evolvable Gene Regulation Gene Regulation Networks Networks

slide-24
SLIDE 24

Backing up again: A Formal Framework for an Backing up again: A Formal Framework for an Operon Operon-Operator Gene Regulation Model

  • Operator Gene Regulation Model

( (Britten Britten-Davidson)

  • Davidson)
  • J. Holland: Adaptation in Natural and Artificial Systems

48

slide-25
SLIDE 25

Broadcast Units Broadcast Units

F F BC[

BC[ S S1

1,

, S S2

2,

, S S3

3,

, S S4

4]

] If If at time at time t t a signal of type a signal of type S S1

1 is present

is present and and no signal of type no signal of type S S2

2 is present,

is present, then then at time at time t t+1 +1 the signal the signal S S3

3 is broadcast

is broadcast and and the signal the signal S S4

4 is deleted at time

is deleted at time t t. .

slide-26
SLIDE 26

Gene Regulation with BC Units Gene Regulation with BC Units

F F Sensor-integrator gene complex

Sensor-integrator gene complex SI SI1

1I

I2

2I

I3

3:

: BC[ BC[ S S, _, { , _, {I I1

1,

, I I2

2,

, I I3

3}, _]

}, _]

F F Receptor-producer complex

Receptor-producer complex R R1

1R

R2

2P

P: : BC[{ BC[{R R1

1,

, R R2

2}, _,

}, _, P P, _] , _]

slide-27
SLIDE 27

Broadcast Language Broadcast Language Example Example

slide-28
SLIDE 28

References References

F Holland, J. H. (1992). Adaptation in

Natural and Artificial Systems. Cambridge, MA, MIT Press.