Krasnow Institute for Advanced Studies -- George Mason University James Albus Senior Fellow Krasnow...

Krasnow Institute for Advanced Studies -- George Mason University

James Albus

Senior FellowKrasnow Institute for Advanced Studies

George Mason University

[email protected]

Reverse Engineeringthe Brain


Outline

What does reverse engineering mean?

An example from visual perception

What is a path to success?

Some neural computational mechanisms


Building computational machines that are functionally equivalent to the brain

in their ability to perceive, think, decide, and act in a purposeful way to achieve goals in complex, uncertain, dynamic, and possibly hostile environments, despite unexpected events and unanticipated obstacles,

while guided by internal values and rules of conduct.

Reverse Engineering the Brain

Functional equivalence

Producing the same input/output behavior


Will require a deep understanding of how thebrain works and what the brain does

How does the brain generate the incredibly complex colorful, dynamic internal representation that

we consciously perceive as external reality?

Reverse Engineering the Brain

How are signals transformed into into symbols?

How are images processed?

How are messages encoded?

How are relationships established and broken?

What are the knowledge data structures?

What are the functional operations?

How is information represented in the brain?

How is computation performed?


Engineering Requires a Scientific Model

Resolution of the model?

• overall system level (central nervous system) • arrays of macro-computational units (e.g., cortical regions) • macro-computational units (e.g., cortical hypercolumns & loops) • micro-computational units (e.g., cortical microcolumns & loops)

• neural clusters (e.g., spinal and midbrain sensory-motor nuclei)

• neurons (elemental computational units) – input/output functions

• synapses (electronic gates, memory elements) – synaptic phenomena

• membrane mechanics (ion channel activity) – molecular phenomena


Computational Mechanisms

Synapse is an electronic gate

Neuron is an atomic computational element

Cortical Computational Unit is a collection offunctional elements – capable of:

focus of attention, segmentation and grouping,calculation of group attributes and state,classification, and establishing relationships

Neural Cluster is a functional element – capable of:arithmetic or logical operations, correlation, coordinate transformation, finite-state automata, grammar, direct and indirect addressing


A Typical Neural Cluster

e.g. in the cerebellum (Marr 1969, Albus 1971),

S(t) P(t + t) = H(S(t))

memory recall, arithmetic or logical functions,IF/THEN rules, goal-seeking reactive control,

inverse kinematics, direct & indirect addressing


Neural Clusters in Spinal Cord


differential and integral functions, dynamic models,time and frequency analysis, phase-lock loops

S(t) P(t + t) = H(S(t))

A Neural Cluster + Feedback


A Neural Finite State Automaton

Markov processes, scripts, plans, behaviors, grammars

S(t) P(t + t) = H(S(t))

State Next state


Cortical Columns

Microcolumns 100 – 250 neurons 30 – 50 diameter, 3000 longe.g., detect patterns, compute pattern attributes

Hypercolumns (a.k.a. columns)100+ microcolumns in a bundle500 in diameter, 3000 longBasis of Cortical Computational Unit (CCU)

There are about 106 hypercolumns in human cortex


Communication

Axon is an active fiber connecting one neuron to others (i.e., publish-subscribe network, bandwidth ~ 500 Hz)

Two kinds of axons:• Drivers – Preserve topology and local sign Convey data (attributes) e.g., color, intensity, shape, size, orientation, motion

• Modulators – Don’t preserve topology or local sign Convey context (addresses, pointers)* e.g., select & modify algorithms, establish relationships

Sherman & Guillery 2006*my hypothesis


Example fromVision

Retina

Lateral Geniculate

V1

Leftfieldof view

Rightside ofbrain


Representation of Pixels from the Retina


Similar Representation of Pixels from the Skin


Architecture of Vision

CorticalColumns

in V1 +

LateralGeniculate

in Thalamus

Hypercolumn

Modulator Input

Modulator Output toother cortical regions

Input from lgn

Driver Output to Higher Level & superior colliculus

Modulator OutputBack to lgn

Microcolumn

Receptivefield

DiffuseFibers(Modulators)


Cortical Hypercolumn + Thalamic NucleusCortical Computational Unit (CCU)

Drivers(data)

Modulators(addresses)

CCUOutputs


Cortical Hypercolumn + Thalamic loop

windowing, segmentation, grouping, computing group attributes &state, filtering, classification, setting and breaking relationships

CorticalComputational

Unit(CCU)

drivers = attribute vectors

modulators = address pointers


drivers = attribute vectors

Cortico-Thalamic Loop

123456

t

modulators = address pointers

windowingsegmentation & groupingcompute group attributes

recursive filteringclassification

cortical hypercolumnthalamic

nucleus

A CorticalComputational

Unit(CCU)


Cortico-Thalamic

LoopHierarchy


Receptive Field Hierarchy

Defined by driver neurons flowing up the processing hierarchy


Segmentation& Grouping

Process

Each level detects patterns within its

receptive field in the level below


Grouping Pointers


Grouping Hierarchy

Defined by segmentationand grouping processes

Pointers link symbols to pixels& vice versa

Provide symbolgrounding

Pointers reset everysaccade ~ 150 ms


Cortex is Remarkably Uniform

In frontal cortex, drivers flow downCCUs select goals, set priorities,make plans, and control behavior with intent to achieve goals despite uncertainties

In posterior cortex, drivers flow upCCUs link signals to symbols & vice versa-- from pixels to objects and situations (in space)-- from frequencies to events and episodes (in time)


Behavior Generation

In the frontal cortex, hierarchical arrays of CCUs are capable of:

decision-making, planning, coordinating,& controlling millions of muscle fibers in effective goal-directed adaptive behavior


Cortico-Thalamic

Loop inFrontalCortex

This is aPlanning Loop

SpatialModelof ExternalWorld

PredictedResults

ofPlan

Dynamic Model of Own Body

Desired Goal & Contemplated Plan

Command to Execute Plan

From Kandel & Schwartz2001

Timing& Sync

SelectBestPlan


What is the path to success forreverse engineering the brain?

Pick the right level of resolution

There are 1011 neurons and 1015 synapsesin the brain

Real-time modeling at this resolutionis well beyond current technology

Real-time modeling ::= 20 cycles per second


What is the path to success?

Pick the right level of resolution

There are 106 CCUs in the human cortex

Real-time modeling at this resolutionseems within current technology

There are ~ 104 neurons in a CCU

Real-time modeling at this resolution seems within current technology


State of art supercomputer 3 x 1014 fops

Allocating this to 106 CCUs running at 20 Hzyields 1.5 x 107 fops per CCU per cycle

Estimated communication load of about3 x 105 bytes per second for each CCU, or3 x 1011 bps for full brain model

This appears to be within the state of the art

Computational Estimates


Summary & Conclusions• Reverse engineering the brain requires selecting the right level of resolution, e.g. functional modules and connecting circuitry

• Each CCU consists of -- a frame with attributes and pointers

-- computational processes to maintain it

• Cortical Computational Unit (CCU) is a fundamental functional module in cortex

• Real-time modeling at level of functional modules appears feasible with current supercomputers -- maybe with PC computers in 20 years


Questions?

Krasnow Institute for Advanced Studies -- George Mason University James Albus Senior Fellow Krasnow...

Documents

Transcript of Krasnow Institute for Advanced Studies -- George Mason University James Albus Senior Fellow Krasnow...