11/2008aaai 20081 circuit sharing and the implementation of intelligent systems michael l. anderson...

37
11/2008 AAAI 2008 1 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in Neuroscience and Cognitive Science University of Maryland College Park, MD USA Department of Psychology Franklin & Marshall College Lancaster, PA USA

Upload: ella-bryan

Post on 11-Jan-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

11/2008 AAAI 2008 1

Circuit sharing and the implementation of intelligent systems

Michael L. Anderson

Institute for Advanced Computer StudiesProgram in Neuroscience and Cognitive Science

University of MarylandCollege Park, MD USA

Department of PsychologyFranklin & Marshall College

Lancaster, PA USA

Page 2: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Cognitive Architecture

What is the overall functional architecture of the brain?

11/2008 AAAI 2008 2

Page 3: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Cognitive Architecture

The classical, and still most widely accepted answer:

1.Low-level localization of function

2.High-level localization of domain

11/2008 AAAI 2008 3

Page 4: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

4

Low-level localization of function

Penfield’s Homunculus11/2008 AAAI 2008

Page 5: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

5

High-level localization of

domain

Brodmann map showing functional domains11/2008 AAAI 2008

Page 6: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

More abstractly

11/2008 AAAI 2008 6

1

3

5

2

4

6

Classical c.a. (modularity?) suggests:

• Each brain area has a fixed working

• Each function (and class of functions) is implemented in dedicated neural structures

Page 7: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

As opposed to

11/2008 AAAI 2008 7

Holism (connectionist c.a.?) suggests:

• Each brain area has a flexible working

• Each function (or class of functions) is implemented in overlapping neural structures

1

3

5

2

4

6

Page 8: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

As opposed to

11/2008 AAAI 2008 8

Redeployment suggests:

• Each brain area has a fixed working

• Each function (or class of functions) is implemented in overlapping neural structures

1

3

5

2

4

6

Page 9: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

9

What’s redeployment?

Evolutionary considerations favor a “component re-use” model.

Components evolved for one cognitive function are “exapted” for later uses.

However, the original functionality is not lost—hence “redeployment” rather than exaptation.

11/2008 AAAI 2008

Page 10: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Evolution via redeployment

11/2008 AAAI 2008 10

Page 11: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

11/2008 AAAI 2008 11

Modularity vs. Holism vs. Redeployment

1

3

5

2

4

6

1

3

5

2

4

6

1

3

5

2

4

6

Page 12: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Empirical evidence

• Database of 665 (subtraction-based) imaging experiments in 20 cognitive domains.

• “Functional connectivity” analysis of 472 experiments in 8 cognitive domains (all domains with > 30 experiments).

1211/2008 AAAI 2008

Page 13: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Functional connectivity

1) Choose a spatial segmentation of the brain (we currently use Brodmann areas)

2) Choose an independent variable of interest (cognitive domain)

3) Determine which regions are statistically likely to be co-active, for different levels of the I.V.

1311/2008 AAAI 2008

Page 14: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Step 3 in more detail

A. Calculate chance probability (Q) of co-activation for each BA pair

B. In each domain, determine observed probability (K) of co-activation of each BA pair

C. Where there is a significant difference between Q and K (Χ2), this is considered a “functional connection”

1411/2008 AAAI 2008

Page 15: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Functional cooperation

15

• Functional connection indicates areas that cooperate in service of cognition

11/2008 AAAI 2008

AB -AB

Domain Co-active in domain

Not co-active in domain

-Domain Co-active not in domain

Not co-active not in domain

Page 16: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

List of domains

16

Domain N

Action 56

Attention 77

Emotion 42

Language 165

Memory 88

Mental imagery 31

Reasoning 33

Visual perception 57

11/2008 AAAI 2008

Page 17: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Functional cooperation

We can make graphs of these cooperation links.

1711/2008 AAAI 2008

ActiveArea CoActiveAreaExpected CoactProb

Observed CoactProb

ChiSquare

BA10L BA32L 0.019 0.036 8.34

BA10L BA32R 0.015 0.054 61.30

BA10L BA40L 0.029 0.054 13.77

BA10L BA40R 0.016 0.036 13.82

BA10L BA44L 0.018 0.036 10.77

BA10L BA44R 0.012 0.036 28.86

Page 18: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Action

18

Page 19: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Attention

19

Page 20: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Language

20

Page 21: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Comparing Domain Complexes

2111/2008 AAAI 2008

Can compare many things, for instance:– Node overlap

• Indicates B.A.s shared by different domain complexes

– Edge overlap• Indicates functional connectivity/cooperation

shared by different domain complexes

– Network topology• May give clues about nature of function

implementation

Page 22: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Node vs. Edge Overlap

Use Dice’s coefficient: 2(o1,2)/(n1+n2)

Predictions:– Modularity: e, n– Holism: E, N– Redeployment: e, N

2211/2008 AAAI 2008

Page 23: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

11/2008 AAAI 2008 23

Modularity vs. Holism vs. Redeployment

1

3

5

2

4

6

1

3

5

2

4

6

1

3

5

2

4

6

Page 24: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Nodes vs. Edges

11/2008 AAAI 2008 24

Page 25: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Nodes vs. Edges

11/2008 AAAI 2008 25

Page 26: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Nodes vs. Edges

11/2008 AAAI 2008 26

Page 27: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Nodes vs. Edges

11/2008 AAAI 2008 27

p << 0.001

Page 28: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

But . . .

Maybe this result is just an artifact

• Given a small number of nodes (84)

• Large number of possible edges (3486)

• Get high node overlap and low edge overlap just by chance

11/2008 AAAI 2008 28

Page 29: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

11/2008 AAAI 2008 29

p << 0.001

Page 30: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

30

4 implications1. Give up on modularity in its classic form

2. Need to develop a domain-neutral vocabulary for cognitive science

3. Assigning computational/cognitive roles to brain areas will require cross-domain modeling

4. Should consider cross-domain uses when designing cognitive components

11/2008 AAAI 2008

Page 31: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

31

4 implications1. Give up on modularity in its classic form

2. Need to develop a domain-neutral vocabulary for cognitive science

3. Assigning computational/cognitive roles to brain areas will require cross-domain modeling

4. Should consider cross-domain uses when designing cognitive components

11/2008 AAAI 2008

Page 32: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Divergence in implementation

11/2008 AAAI 2008 32

Modular architectures support functional assignment by decomposition and analysis

Page 33: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

Convergence in implementation

11/2008 AAAI 2008 33

Complex system

Functional Complex 2

Functional Complex 3

Component 1

Functional Complex 1

Component 2

Component 3

Component 4

Component 5

Component 6

Sub-component 5

Sub-component 4

Sub-component 3

Sub-component 2

Sub-component 6 . . .

Sub-component 1

Page 34: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

34

Cross-domain modeling

1. Cannot determine what a sub-component should do by considering only an individual task or task category, as been the normal practice.

2. Must begin to consider at design time the use of low-level components across multiple tasks in multiple domains.

11/2008 AAAI 2008

Page 35: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

35

Cross-domain modeling (2)

To do this:

1. Model each function of the system

2. Map sub-functions to a limited set of components

3. Constraint: each point of overlap must assign same (abstract) sub-function to each component

11/2008 AAAI 2008

Page 36: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

36

Cross-domain modeling (3)

11/2008 AAAI 2008

Page 37: 11/2008AAAI 20081 Circuit sharing and the implementation of intelligent systems Michael L. Anderson Institute for Advanced Computer Studies Program in

• Anderson, M.L. (2007). The massive redeployment hypothesis and the functional topography of the brain. Philosophical Psychology, 21(2): 143-174.

• Anderson, M.L. (2007). Evolution of cognitive function via redeployment of brain areas. The Neuroscientist, 13(1): 13-21.

• Anderson, M. L. (2007). Massive redeployment, exaptation, and the functional integration of cognitive operations. Synthese 159(3): 329-45.

• Anderson, M.L. (2008). Circuit sharing and the implementation of intelligent systems. Connection Science, 20(4): 239-51.

http://www.agcognition.org

11/2008 AAAI 2008 37