the theory and practice of computational cognitive neuroscience

49
Dr. Brian J. Spiering email: [email protected] website: brianspiering.com The Theory & Practice of Computational Cognitive Neuroscience

Upload: brianspiering

Post on 20-May-2015

237 views

Category:

Technology


1 download

DESCRIPTION

"A call to arms" for Computational Cognitive Neuroscience.

TRANSCRIPT

Page 1: The theory and practice of computational cognitive neuroscience

Dr. Brian J. Spieringemail: [email protected]

website: brianspiering.com

The Theory & Practice of

Computational Cognitive Neuroscience

Page 2: The theory and practice of computational cognitive neuroscience
Page 3: The theory and practice of computational cognitive neuroscience
Page 4: The theory and practice of computational cognitive neuroscience
Page 5: The theory and practice of computational cognitive neuroscience
Page 6: The theory and practice of computational cognitive neuroscience

Oblate Spheroid

Page 7: The theory and practice of computational cognitive neuroscience

Oblate Spheroidwith unequal mass distribution

Page 8: The theory and practice of computational cognitive neuroscience

Choose the modelbased on the task

Page 9: The theory and practice of computational cognitive neuroscience

OverviewI) Context

II) Theory

III) Practice

Page 10: The theory and practice of computational cognitive neuroscience

1) ContextA) How to Buy a Science model

B) Validating Models

C) Hierarchy of Models

D) The First Step

Page 11: The theory and practice of computational cognitive neuroscience

What to look for when “buying” a science model ...

1) Accounts for previously accounted for data

2) Accounts for previously unaccounted for data

3) Makes novel, falsifiable predictions

Page 12: The theory and practice of computational cognitive neuroscience

Validating Models

It is easy to account for behavioral data

so add constraints ...

- wider (more tasks)

- deeper (more methodologies)

- neuroscience

Page 13: The theory and practice of computational cognitive neuroscience

Hierarchy of Models

No Model

Modules

Circles & Lines

CCN

Page 14: The theory and practice of computational cognitive neuroscience

Why is CCN on the top?

Can predict in these domains:

- Behavioral

- fMRI

- MEG & EEG

- Single-unit recording

Page 15: The theory and practice of computational cognitive neuroscience

Why is CCN on the top?Can predict account for these factors:

- Experimental Manipulation

- Drugs

- Disease

- Genes

- Focal Lesions (injury, ablation, & TMS)

Page 16: The theory and practice of computational cognitive neuroscience

Where do you start?

Model ➜ Brain

or

Brain ➜ Model

Page 17: The theory and practice of computational cognitive neuroscience

II) Theory of ModelingA) Ideals

B) Modeling individual units

C) Learning

D) Behavior

E) Other Issues

Page 18: The theory and practice of computational cognitive neuroscience

A) Ideals1) Neuroscience

2) Simplicity

3) Set-in-Stone

4) Goodness-of-Fit

Page 19: The theory and practice of computational cognitive neuroscience

1) Neurosciencea) Use only found connections

b) Specify excitatory or inhibitory

c) Account for qualitative behavior in units

d) Agree with neuroscience of learning

Page 20: The theory and practice of computational cognitive neuroscience

2) Simplicity

Use enough but not any more

Page 21: The theory and practice of computational cognitive neuroscience

3) Set-in-Stone

Always use the same players and play by

the same rules in the modeling game.

Page 22: The theory and practice of computational cognitive neuroscience

4) Goodness-of-Fit

a) Behavioral

b) Neuroscience

Page 23: The theory and practice of computational cognitive neuroscience

B) Modeling Units1) Leaky integrate-and-fire model

2) Izhikevich model

3) Axon & synaptic delays

Page 24: The theory and practice of computational cognitive neuroscience

C) Learning1) LTP vs. LTD

2) Role of Dopamine

3) Discrete vs. Continuous

4) Local vs. Global

Page 25: The theory and practice of computational cognitive neuroscience

D) Behavior

1) Choose control region

2) Choose function

Page 26: The theory and practice of computational cognitive neuroscience

E) Other Issues1) Overconnectioning

2) Other?

Page 27: The theory and practice of computational cognitive neuroscience

III) Practice of ModelingSPEED Model:

A) Idea

B) Inception

C) Instantiations

Page 28: The theory and practice of computational cognitive neuroscience

A) SPEED Model IdeaAn Contrarian Model of Procedural Expertise

F. Greg Ashby John Ennis

Page 29: The theory and practice of computational cognitive neuroscience

SPEEDAn Contrarian Model of Procedural Expertise

Page 30: The theory and practice of computational cognitive neuroscience

Previous Notions

“It has been widely held that although memory traces are at first formed in the cerebral cortex, they are finally reduced or transferred by long practice to subcortical levels” (p. 466)

Karl Lashley (1950) In search of the engram.

Page 31: The theory and practice of computational cognitive neuroscience

Previous Notions

“Routine, automatic, or overlearned behavioral sequences, however complex, do not engage the PFC and may be entirely organized in subcortical structures” (p. 323)

Joaquin Fuster (2001). The prefrontal cortex – an update.

Page 32: The theory and practice of computational cognitive neuroscience

SPEEDAn Contrarian Model of Procedural Expertise

Page 33: The theory and practice of computational cognitive neuroscience

SPEED Model

Procedural learning is striatum dependent.

Procedural expertise is striatum independent.

SPEED (Subcortical Pathways Enable Expertise Development)

Page 34: The theory and practice of computational cognitive neuroscience

B) Inception

aka, How we built the model.

Page 35: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

Response

BA

B

A-B difference

A B

A BA

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:

Page 36: The theory and practice of computational cognitive neuroscience

SensoryAssociationA Response

A

A

AA

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

ExcitatoryInhibitory

SPEED Model:Single Subcortical

Page 37: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

Response

BA

B

A-B difference

A B

A BA

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:Subcortical

Page 38: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

StriatumSNpc

Dopamine

Excitatory

SPEED Model:Subcortical Learning

Page 39: The theory and practice of computational cognitive neuroscience
Page 40: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

Response

BA

B

A-B difference

A B

A BA

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:Subcortical

Page 41: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

Response

BA

A-B difference

SPEED Model:Cortical

Page 42: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

Response

BA

B

A-B difference

A B

A BA

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:Cortical Learning

Page 43: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

A

A

A A

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:1st Trial

Page 44: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

A

A

A A

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:Early Learning

Page 45: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

A

A

A A

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:Middle Learning

Page 46: The theory and practice of computational cognitive neuroscience

SensoryAssociation

Cortex

A

A

A A

Premotor Area(Cortex)

Thalamus

Globus PallidusStriatum

SPEED Model:Mature Performance

Page 47: The theory and practice of computational cognitive neuroscience

C) Instantiations

See MATLAB code ...

Page 48: The theory and practice of computational cognitive neuroscience

Summary1) Context

II) Theory

III) Practice

Page 49: The theory and practice of computational cognitive neuroscience

?