Design principles for adaptive self-organizing systems Finding Fluid Form Symposium University of Brighton December 9-10, 2005 Peter Cariani www.cariani.com Department of Physiology Tufts Medical School Boston
Mar 26, 2015
Design principles for
adaptive self-organizing systems
Finding Fluid Form SymposiumUniversity of Brighton
December 9-10, 2005
Peter Cariani
www.cariani.comDepartment of Physiology
Tufts Medical School
Boston
Organismic biology (undergrad @ MIT mid 1970s)
Biological cybernetics & epistemology (1980s)Biological alternatives to symbolic AI
Howard Pattee, Systems Science, SUNY-Binghamton
Temporal coding of pitch & timbre (1990s)
Auditory neurophysiology, neurocomputationHow is information represented in brains?
Commonalities of coding across modality & phyla Neural timing nets for temporal processing
Auditory scene analysisPossibilities inherent in time codesTemporal alternatives to connectionism
signal multiplexing; adaptive signal creation broadcast
My trajectory
Elaboration of structures & functions over time in biological, social, and technological realms,
What makes new functions possible (functional emergence)? Can we put these principles to work for us?Is structural complexification by itself sufficient? (No)Notions of function & functional emergence are needed.What kinds of functions? Sensing, effecting, coordinatingIs pure computation on symbols sufficient? (No)How are brains/minds capable of open-ended creativity?Neural codes, temporal codes, timing nets Neural coding of pitch in the auditory systemRethinking the architecture of the brain:
Temporal alternatives to connectionismAdaptive signal creation & multiplexing, Broadcast coordinative strategies
Evolution of ideas
D
F1
5
1-2-F
M
D-F-1
2-F-M
M-D
1-M
Processes forcombining primitives
Combinatoric emergence:New combinations
of pre-existing primitives
Creative emergence:De novo creation of new primitives
DF 1
5
1-2-F
M
D-F-1
2-F-*
M-D
a-M
a *
Process forconstructing
newprimitives
Sets ofpossible
combinations of primitives 2- -F M
Sets of primitives( , , )axioms atoms states
Add *, a
Combinatoric vs. creative emergence
An example
Exhaustive description Limited description
All permutations of
6 arbitrarily defined objects
All permutations of
single digits
0 1 2 3 4 5 6 7 8 9
consisting of 6 tokens
One well-defined set
having 610 permutations
BOUNDED
Ill-defined number of sets, each w. 610 permutations
UNBOUNDED
Describing the world: Two perspectives
Omniscent
“God’s eye view”
Postulational,
ontological
analytical mode
Perspective of
the limited observer
epistemological
empirical mode
Appearance of newstructures over time
Violations of expectations“Surprise”
Well-defined vs. ill-defined realms
Exhaustive description
God’s eye view
Limited description
Limited observer
Description of
all-possible
organism-environment
relations
Environment as
ill-defined realm
Description is
dependent on set of observables
(environment has as many properties as one can measure)
System-environment as
well-defined realm
No fundamental novelty is possibleAll novelty is combinatoric
Combinatoric andCreative emergence
CLOSED WORLD ASSUMPTION OPEN WORLD ASSUMPTION
A
F e a t u r e 1
A
A
B
B
B
T h r e e f e a t u r e sl i n e a r l y
s e p a r a b l e
A
B
B
A
F e a t u r e 1
A
A
A
A
B
B
B
B
A
B
T w o f e a t u r e sn o t l i n e a r l y
s e p a r a b l e
Effect of adding a new observable
New features
CREATING A NEW OBSERVABLE ADDS A NEW PRIMITIVETHAT INCREASES THE EFFECTIVE DIMENSIONALITY OF THE SYSTEM
OntologyAristotelian hylomorphism
Material substrate that exists independently of us,
yet whose form is largely ill-defined, incompletely known
Organization is embedded in material system
(e.g. mind is the organization of the nervous system)
Conscious awareness requires a particular kind of
regenerative informational organization
embedded in a material system (cybernetic
functionalism)
Aristotle's Causes: Multiple complementary modes of
explanation that answer different kinds of questions
Philosophy
These are different modesof explanation that answerdifferent kinds of questionsabout the behavior of thesystem.
Functional organizationand form are embedded inthe material substrate;consequently, they changewhen the substrate changes.
Mind is the functionalorganization of the brain.
The structure of conscious experiencemirrors the structure of the mind.
Epistemology
Pragmatism (truth of a model related to its purpose)
Perspective of the limited observer
Relativism: different observational frames & purposes
Analytical, empirical and pragmatic truths
Analytic: truths of convention (non-material truths, finist mathematics)
Empirical: truths of measurement, observation (science)
Pragmatic: truths of efficacy & aesthetics (engineering, art)
Constructivism & epistemic autonomy:
by semi-freely choosing our own observables & concepts,
we construct ourselves (for better or worse)
Philosophy
We are interested in
designing & fabricating systems that
autonomously organize themselves
to elaborate structures & improve functions
in response to challenges of their environments
in ways that are meaningful and useful
to us and/or them
Design principles for adaptive, self-organizing systems
Richness of material possibility (e.g. polymeric combinatorics)
+ Ability to steer & stabilize structure(feedback to structure: sensors, coordination mechanisms, effectors)
+ Means to interact w. material world
(sensing, action = "situatedness", semantics)
+ Means to evaluate actions re: purposes(goal-laden representations, "intentionality")
---------------------------------------------------------------------------------------
-
=> Material system capable of adaptive,
elaboration & improvement of informational
functions
Design principles for adaptive, self-organizing systems
Richness of material possibility (need polymers, replicated aperiodic structure, Schrodinger's aperiodic crystal,
analog dynamics, ill-defined interactions)
Ability to steer & stabilize structure(need controls on self-production of internal structure, enzymes)
Means to interact w. material world
(Need sensors, effectors, neural nets)
Means to evaluate actions re: purposes
(Need natural selection or internal goal states, limbic system)
Design principles for adaptive, self-organizing systems
QuickTime™ and aUSBVision decompressor
are needed to see this picture.
Richness of material possibility
Complexity is easy
Steerable complexity is hard
QuickTime™ and aUSBVision decompressor
are needed to see this picture.
Vibratory dynamics of matter
Cymatics:
Bringing Matter to Life
with Sound
Hans Jenny
Material possibility+ Steer, stabilize, specify, inherit
+ Sensorimotor interaction + Evaluation => ASOS
Design principles for adaptive, self-organizing systems
VARIATION + SELECTION + INHERITANCE => ADAPTATION
Two phases in creative learning processes
Expansive phase: generation of possibilityRealm of free & open creation
e.g. scientific imagination and hypothesis creation Contractive phase: selection of best possibilitiesRealm of clarity & rigorous evaluation
e.g. hypothesis testing (clarity, removal of ambiguity)
Analog dynamics and discrete symbols
We will also argue that one almost inevitably needs
mixed analog-digital systems for complex systems:
i.e. systems w. analog dynamics constrained by digital states ("symbols")
for reliable replication of function
for inheritability of adaptive improvements
Analog and digital are complementary modes of description
analog descriptions - continuous differential equations
digital descriptions - discrete states & ST rules/probabilities
Digital states or discrete symbols are ultrastable basins of attraction
Different theoretical approaches tounderstanding brains and their functions
Dynamicalsystems
approachesSymbol-
processing
Neuralinformationprocessing
differential growthhomeostasisanalog
states & switchesbranchingdiscrete
representationsprocessing
Requisite: sensorimotor loopsInner and outer loops
metabolism: self-productionsteering: percept-action coordinations
action
perception
interaction w.environment
Von Uexküll’s umwelts
McCulloch’sinternal andexternal loops
.
sensors
alterstructure
evaluateselect
Pragmatics
Syntactics
environment
coordinationPerceptstates
Decisionstates
effectors
Action
Performance
Predicted result
Observed result
World2
measure
Sp
So
World1
InitialConditions Predictive model
Physical laws
Formal rules
Encoding(Semantic)
(Syntactic)
Observer'schoices:
what to predictwhat to measure
(Pragmatic)
measure
Realm ofsymbolic
description
Realm ofmaterialaction
Si
A
Self-conscious description of the modeling process:Hertzian modeling relation: measurement & computation
The choice of observables
Finding the variablesThe would-be model maker is now in the extremely common situation of facing some incompletely defined "system," that he proposes to study through a study of "its variables." Then comes the problem: of the infinity of variables available in this universe, which subset shall he take? What methods can he use for selecting them?
W. Ross Ashby, "Analysis of the system to be modeled" in: The Process of Model-Building in the Behavioral Sciences, Ohio State Press, pp. 94-114; reprinted in Conant, ed. Mechanisms of Intelligence
The choice of observables - analogous problems
1. Choice of primitive features for classifiers
2. Evolution of sensory organs in organisms
3. Choice of sensors for robots
A
F e a t u r e 1
A
A
B
B
B
T h r e e f e a t u r e sl i n e a r l y
s e p a r a b l e
A
B
B
A
F e a t u r e 1
A
A
A
A
B
B
B
B
A
B
T w o f e a t u r e sn o t l i n e a r l y
s e p a r a b l e
Effect of adding a new observable
alterstructure
measure evaluate
Syntactic Axis
Pragmatic Axis
control
Computation
S i Sf
environment
Semioticsof adaptivedevices
Feedback to state
Feedback to structure alters functionalities
Semiotic relations (Charles Morris)
SYMBOL(Functional state)
SYSTEM-GOALS EXTERNALWORLD
OTHERSYMBOLS
Semanticspercept-action linkagesPragmatics
valuations
Syntacticsrules on symbol - types
PURPOSE
MEANING
Sensory inputsMotor outputs
WORLD
BRAIN
Externalsemanticlinkages
Syntacticlinkages
Pragmaticlinkages
DESIRESDRIVES THOUGHTS
SENSATIONS
B
Evaluatere: goalsFrontal & limbic systems
Internallygeneratedpatternsequences
sensorysystems
motorsystems
Adaptivity in
percept-action
loops (Cariani)
∆ = alter structure → alter function
evaluate
Syntactic Axis
Si S
f
∆∆control
environment
SemanticAxis
PragmaticAxis
measure
∆
coordination
C
symbolic
nonsymbolic
environment
user-specified initial state
Computation
final state
state trans. rules
Pure computation (state-determined system, no independent informational transactions w. environment)
m e a s u r e c o n t r o l
sensorarray effector
array
a c t i o n
v e c t o r
f e a t u r e
v e c t o r
performance
e n v i r o n m e n t
computation
Fixed robotic device
Fixed sensors,coordinators, and effectors;
Purely reactiveand driven by its inputs;
Incapable oflearning
Computationallyadaptivedevice
computation
a c t i o n
v e c t o r
f e a t u r e
v e c t o r
∆
performance
training
c o n t r o lm e a s u r e
e n v i r o n m e n t
t e s t
Trainable machinesNeural networksAdaptive classifiersGenetic algorithmsRobots w. adaptive programs
Capable of learningnew percept-actionmappings (classifications)
Some observations about adaptability
Whatever functionalities are fixed, the designer must specify
works for well-defined problems & solutions
advantage: predictable, reliable behavior
drawback: problems of specification
Whatever is made adaptive must undergo a learning phase
needed for ill-defined problems & solutions
some unpredictability of solutions found
creative behavior!
the more autonomy, the more potentially creative
Consequently, there are tradeoffs between
adaptability & efficiency
autonomy/creativity & control/predictability
coordinationfeature vector
test
performance
adaptiveconstruction
of a new sensor∆sensors
action vector
environment
Evolution/adaptive construction of new sensors
sensory evolutionimmune systemsperceptual learning
capable of learningnew perceptual categories
new feature primitives
(new observables)
• When a system can choose its own categories – through which it perceives and acts on the world – that system achieves some limited degree of epistemic autonomy.
• A rudimentary electrochemical device was built by cyberneticist Gordon Pask in 1958 that grew its own sensors to create its own “relevance criteria.”
Epistemic autonomy
"With this ability to make or select proper filters on its inputs, such a device explains the central problem of epistemology. The riddles of stimulus equivalence or of local circuit action in the brain remain only as parochial problems." .
Warren McCulloch, preface ,Gordon Pask (1961) . An Approach to Cybernetics.
From: "Physical analogues to the growth of a concept", Symposium onthe Mechanization of Thought Processes, National Physical Laboratories,November 24-28, 1958, H.M.S.O., London, Volume II, p.919.
Principles of action/use
1. Front-ends for trainable classifiers
Useful in ill-defined situations where one does not a priori know what features are adequate to effect a classification
2. Adaptive, self-organizing sensors
Grow structures over analog-VLSI electrode arrays in order to sense new aspects of the world. Use biochemical and/or biological systems coupled to an electrode array
3. Materially-based generator of new behaviors (adaptive pattern-generators)
Similar steerable, ill-defined systems could be used to generate new patterns (sound, images) in an open-ended way that is not at all obvious to the observer/controller
4. Epistemic autonomy
Device chooses how it will be connected to the outside world; what aspects of the material world (categories) are relevant to it. (Symbol grounding, frame problem)
Feedback to state vs. feedback to structure
A thermostat is limited in the information that it can gain from its environment by the fixed nature of its sensors. It has feedback to state, but not feedback to structure. The amount of information that such a system can extract from its environment is finite at any time, and bounded by its fixed structure.
A system capable of sensory evolution or perceptual learning has the ability to change its relation to its environs. Such a system has an open-ended set of observational primitives. It has both feedback to state and feedback to structure. The amount of information that such a system can extract from its environment is finite at any time, but unbounded. Such a system is open-ended.
Analog dynamics without inheritable constraint (Hans Jenny)
QuickTime™ and aUSBVision decompressor
are needed to see this picture.
von Neumann's kinematic (robotic) self-reproducing automaton (1948)
Inheritable constructionanalog dynamics constrained & selected by discrete symbols
∆
constructionpossibilitiesA B C Dα β χ δ1 2 3 4
computationfeature vector
testperformance
∆∆
controlmeasure
action vector
environment
constructionlanguage
constructall parts ofthe device
select fromexisting alternatives
physicalconstruction
Aδ3 (mutation)
Symbolically-encodedmemory permits results ofan optimization process tobe passed to subsequentgenerations
Purely analog adaptivesystem must be trainedeach generation
Genetic algorithm +Pattern grammar forguiding constructionconstrained search
The homeostat
Relation to Ashby's homeostat
∆
constructionpossibilitiesA B C Dα β χ δ1 2 3 4
computationfeature vector
testperformance
∆∆
controlmeasure
action vector
environment
constructionlanguage
constructall parts ofthe device
select fromexisting alternatives
physicalconstruction
Aδ3 (mutation)
Evaluation of ability to control inputs
Analog sensor/controller
Uniselector25x25x25x25 = 390k construction possibilities -> variety of the control system, unconstrained search
Relation to Ashby's homeostat
∆
constructionpossibilitiesA B C Dα β χ δ1 2 3 4
computationfeature vector
testperformance
∆∆
controlmeasure
action vector
environment
constructionlanguage
constructall parts ofthe device
select fromexisting alternatives
physicalconstruction
Aδ3 (mutation)
Evaluation of ability to control inputs
Analog sensor/controller
Uniselector25x25x25x25 = 390k construction possibilities -> variety of the control system, unconstrained search
Ashby's homeostat
Environment
Analog controller
(ill-defined structure)
Uniselector
evaluate(in bounds?)
Adaptive analogcontroller
Structure of particular controllersis unknown to designer
Requisite variety forcontrol is the number ofalternative controllersavailable25x25x25 = 390,625
The homeostat & the brainA few cybernetics-inspired accounts of brain function
Sommerhoff (1974) Logic of the Living Brain
Klopf, The Selfish Neuron
Arbib, The Metaphorical Brain
Most successful neuroscientific application of cybernetics:
W.Reichardt's analysis of fly optomotor loop
The homeostat never caught on as a brain metaphor
Some possible reasons:• Homeostats never were cast in terms of neural nets• No obvious digital uniselector function in the brain• Predominance of problems of pattern recognition and
formulation of coherent action over simple problems of internal regulation
The brain as an adaptive self-organizing system
Ideas that flow from cybernetics and theoretical biology:
1) Brains as signal self-production systems
related to reverberant loops (a la Lorente, Lashley, Hebb, McCulloch, Pitts & many others)
2) Brains as pattern-resonance systems
related to Lashley, Hebb, many others
3) Brains as multiplexed signaling and storage systems
holographic paradigms, Longuet-Higgins, Pribram,John
4) Brains as mass-dynamics, broadcast systems
5) Brains as communications nets that create new signals
6) Brains as temporally-coded pulse pattern systems
I believe all this is possible using temporal pattern codes.
Regeneration of parts
F(A)F(B)F(C)
A+B+CF(D)
geneticconstruction
plans
geneticexpressionapparatus(universal
constructor)
replication of plans
replicationof constructor
plans
byproduct D
effect on survivalof whole system
A
byproductsrawmaterials
metabolicloops
B
byproductsrawmaterials
geneticcontrol
geneticexpression &reproduction
genetic plans(symbolic memory)
set boundaryconditions
C
Von Neumann’s kinematic self-reproducing automaton
F(A)F(B)F(C)
A+B+CF(D)
geneticconstruction
plans
geneticexpressionapparatus(universal
constructor)
replication of plans
replicationof constructor
plans
byproduct D
effect on survivalof whole system
A
Autopoiesis and autocatalysis
byproductsrawmaterials
metabolicloops
B
Symbolically-guided self-production
byproductsrawmaterials
geneticcontrol
geneticexpression &reproduction
genetic plans(symbolic memory)
set boundaryconditions
C
Autopoiesis and autocatalysisLife is built upon cycles of self-production
byproductsrawmaterials
metabolicloops
B
Brain function may be based on self-productions of spike patternsHebbian reverberant eigenstates and regenerative temporal patterns
A B
Resonantstate | A
Resonantstate | B
Contingentsensory input
Epistemiccut
S2
S3
S1
R2
R3
R1
CNS
Potentialstimuli
Potentialresponses
A BStimulus-driven switching betweenreverberant circuits (after Hebb, 1965)
Reverberant patternsas switchable eigenstates
Motorresponse
for A
Motorresponse
for B
McCulloch & Pitts (1943) Nets with circles render activity independentof time and semi-autonomous re: the environmentvon Foerster (1948) brain eigenstates as a form of ST memory
Why the mind is in the headWarren McCullochL.A. Jeffress, ed. Cerebral Mechanisms of Behavior(The Hixon Symposium, Wiley, 1951, reprinted inEmbodiments of Mind, MIT, 1965, concluding lines)
This brings us back to what I believe is the answer to the question: Why is the mind in the head?
Because there, and only there, are hosts of possible connections to be formed as time and circumstance demand. Each new connection serves to set the stage for others yet to come and better fitted to adapt us to the world, for through the cortex pass the greatest inverse feedbacks whose function is the purposive life of the human intellect.
The joy of creating ideals, new and eternal, in and of a world, old and temporal, robots have it not.
For this my Mother bore me.
The brain as aself-regeneratingpattern-resonancesystem
The same [resonance] is true of all bodies which can yield notes. Tumblers resound when a piano is played, on the striking of certain notes, and so do window panes. Nor is the phenomenon without analogy in different provinces. Take a dog that answers to the name "Nero." He lies under your table. You speak of Domitian, Vespasian, and Marcus Aurelius Antonius, you call upon all the Roman Emperors that occur to you, but the dog does not stir, although a slight tremor of his ear tells you of a faint response of his consciousness. But the moment you call "Nero" he jumps joyfully towards you. The tuning fork is like your dog. It answers to the name A.
Ernst Mach, Popular Lectures, “The fibers of Corti” c. 1865
Tuning in nervous systemsMinds as pattern-resonances
Pattern resonances: neural assemblies emitting annotative tag signals that elaborate a regenerating signal pattern
Primaryinteractions
Secondaryinteractions
Higher-order, morecomplex interactions
Figure 7. Time-coded broadcast schema for asynchronous, heterarchical global integration.
Incoming sensory signals
Creation of new primitive time patterns
Higher-orderinteractions
Miller, Ratliff, and Hartline (1961) How cells receive stimuli.Scientific American 215(3):222-238.
Phase-locking in visual neurons(Horseshoe crab ommatidium, 5-15 Hz flashes)
Phase-locking in auditoryl neuronsCat auditory nerve fibers, 250 Hz tone
Javel,
Higher-order interval pattern
Multiple intervals in same spike train
Interspike interval code
Temporalpatterncodes
Temporal pattern codes
250 Hz tone
Javel E, McGee JA, Horst W, Farley GR, Temporal mechanisms in auditory stimulus coding.In: G. M. Edelman, W. E. Gall and W. M. Cowan, ed, Auditory Function: NeurobiologicalBases of Hearing, Wiley: New York 1988; p. 518.
Phase-locking in auditory nerve fibers
Frequency and time in the auditory nerve
.
0
50
100
.1 1 10
Cochlea
Frequency(kHz)
Thresholdsound pressurelevel (dB SPL)
Threshold tuning curvesfor discharge rate
1
Peristimulus time (ms)
0 5 10 15 20 25
Stimulus waveform
Peristimulus time histograms
(100 presentations at 60 dB SPL)
Fundamental period 1/F0Formant period 1/F1
Single-formant vowel
10
.
Peristimulus time (ms)
10
1
0 5010 20 30 40
Phase-locking of discharges in the auditory nerveCat, 100x @ 60 dB SPL
Temporal codingin the auditory nerve
Work with Bertrand DelgutteCariani & Delgutte (1996)
Dial-anesthetized cats.100 presentations/fiber60 dB SPL
Population-interval distributions are compiled by summing together intervals from all auditory nerve fibers.
The most common intervals present in the auditory nerve are invariably related to the pitches heard at the fundamentals of harmonic complexes.
Est. Error: 1.9%Est. Error: 0.4%
1000 50 10050
Interspike interval (ms) Interspike interval (ms)
16 Hz5743 intervals
32 Hz467 intervals
Temporal period estimation for an LGN unit(sinusoidal luminance modulation)
20
40
60Rate tuning curve
Frequency (Hz)10 10010.1
Rate(spikes/s)
Raw spike train data courtesy of Andrzej Przybyszewski & Dan Pollen
Phase-locking in visual thalamus (LGN)
Stimuli:Drifting sinusoidalgratings
A B
0 150
red
magenta
cyan
green
yellow
blue
neutral
Time (ms)
5000trolands
• The psychophysical effect.A. Rotating disk patterns used by Fechner,Helmholtz, and Benham to induce subjective colors.B. Temporally-modulated illumination patternsproduce characteristic flicker colors (Festinger, Allyn& White, 1973).
• Electrical stimulation of the retina using similarpatterns produces phosphenes of the correspondingcolors (Young, 1977).
• Wavelength-dependent temporal dischargepatterns have been found in the optic nerve and inhigher visual centers by Granit(1943), Kozak &Reitboeck (1974), and others.
• Implications for neural coding. The existence ofthe psychophysical effect for both natural andelectrical stimulation and its correlate in temporaldischarge patterns of visual neurons suggests thepossibility of a central temporal code for color.
A temporal code for color?
Color vision
NTS
Chordatympani
NTS
Chordatympani
NaCl Quinine HCl
HClSucrose
Temporal coding of taste
Figure 12. Typical temporal patterns of response to four different stimuli recordedfrom the chorda typani nerve and the NTS. The stimuli were: 0.1M NaCl, 0.1Mquinine HCl, 0.5M sucrose and 0.1M HCl. Time markers indicate 5 sec intervals. Thechorda tympani recordings are from the whole nerve, NTS recordings are from singleneurons. The tracings of relative spike frequency shown in this figure were obtainedduring neural recording sessions by the output of amplified neural activity through aspike amplitude window discriminator to a counting rate meter, the output of whichwas displayed on a Brush pen writer. From: Ellen Covey, Temporal Neural Coding of Gustation (1980), Ph.D. thesis, Duke University.
NTS
Chordatympani
NTS
Chordatympani
NaCl Quinine HCl
HClSucrose
Temporal coding of taste
Figure 12. Typical temporal patterns of response to four different stimuli recordedfrom the chorda typani nerve and the NTS. The stimuli were: 0.1M NaCl, 0.1Mquinine HCl, 0.5M sucrose and 0.1M HCl. Time markers indicate 5 sec intervals. Thechorda tympani recordings are from the whole nerve, NTS recordings are from singleneurons. The tracings of relative spike frequency shown in this figure were obtainedduring neural recording sessions by the output of amplified neural activity through aspike amplitude window discriminator to a counting rate meter, the output of whichwas displayed on a Brush pen writer. From: Ellen Covey, Temporal Neural Coding of Gustation (1980), Ph.D. thesis, Duke University.
Temporal codingof taste
time
A1 B1 C1 D1 A2 B2 C2 D2 E2
cycle 2, slots A-Ecycle 1, slots A-E
E1
A. Time-division multiplexing in telephony (signals A-E)
B. Time-division multiplexing in neural systems
time
Scanning Synchrony
Sortingbysynchronousactivation
Featuredetector
units(neural
channels)
time
D. Code-division multiplexing in neural systems
AZ1:9808 DX1:8708 CV2:071389 DX2: 708773CV1:98458 BY2:8708 CV3:134389 AZ4:70767887
BY1:438074 AZ2:4329633 EW3:23342345EW1:43741541 BY3:432963323 EW2:2334245
C. Code-division multiplexing in telephony or computer networks
Sortingbyheader
ReceiversW-Z
Senders A-ESegments1-4
Sortingbytimepattern
Neural assemblyreceivers
Five primitivetime patterns
Neural timing nets
FEED-FORWARD TIMING NETS• Temporal sieves• Extract (embedded) similarities• Multiply autocorrelations
RECURRENT TIMING NETS• Build up pattern invariances• Detect periodic patterns• Separate auditory objects
Relative delay τ
Timet
Sj(t)
Si(t)
Si(t) Sj(t - τ)
individualmultiplicative
term
Si(tm) Sj(tm - t)Στ
convolutiontime-series
term m
two setsof input
spike trains
In p u t t im e s e q u e n c e
A l l t im e d e la y s p r e s e n t
τ0
τ1
τ2
τ3
C o in c id en c e
u n its
D irec t in p u ts
R ec u r ren t ,
in d irec t in p u ts
T im e p attern s rev erb erate
th ro u g h d elay lo o p s
Potential advantages of temporal pattern pulse codes & timing nets
• Multiplexed signal transmission
• Orthogonality of patterns; less interference
• Flexible multimodal integration
• Encoding of signal identity in itself (logical type)
• Liberate signals from wires
• Broadcast of signals + selective reception
• Nonlocal computational operations
• Mass action (statistical representations)
• Open-ended creation of new signal primitives
In the image of the digital computer, we conceptualize brains as distributed logic machines.
However, temporal correlation machines may prove to be a better metaphor.Temporal expectancies in perceptionTemporal patterning of body processesTemporal structure of movementTemporal expectations and reward structure
(dopamine system, conditioning) Temporal memory traces
Music may have the profound effects that it does because 1) it directly impresses its temporal structure on the activity of many neuronal populations, and 2) the neural codes & computations underlying experience are inherently temporal.
Music, brain, and time
Andy Partridge, xtc
Structural complexity alone is not sufficientPure computation alone is not sufficient
RequisitesSensors & effectorsMixed digital-analog designFeedback to structure, self-productionInheritable, replicable (digital) plansCombinatorics of digital stringsRich analog, ill-defined dynamicsGoal states and steering/selection mechanisms
Possibility of brain as temporally-coded self-organizing system
Conclusions Design principles for self-organizing systems
Temporal coding of sensory information.
Interspike interval (ms)
Pitch period
Peristimulus time (ms)
Pitch period
0 255 10 15 20
From cochlea to cortex
Primaryauditory cortex
(Auditory forebrain)
Auditory thalamus
Inferior colliculus(Auditory midbrain)
Auditory brainstem
Cochlea
Auditory nerve (VIII)
Lateral lemniscus
# spikes
Phase-locking to a 300 Hz pure tone
Evans, 1982
Period histogram (1100 Hz)
First-order interval histogram (1500 Hz)
.
Peristimulus time (ms)
10
1
0 5010 20 30 40
Auditory nerve
LowCF
HighCF
F0
F1
F2
F3
Peristimulus time (ms)
Time domain analysis of auditory-nerve fiber firing rates.Hugh Secker-Walker & Campbell Searle, J. Acoust. Soc. 88(3), 1990Neural responses to /da/ @ 69 dB SPL from Miller and Sachs (1983)
Vowel Formant Regions
Populationinterval
distribution
Interspikeinterval (ms)
150
Six stimuli that produce a low pitch at 160 Hz
Pure tone 160 Hz
AM tone Fm:160 Hz
Fc:640 Hz
Harms 6-12
AM tone
Click train
AM noise
F0 :160 Hz
Fm :160 Hz
Fc:6400 Hz
Fm :160 Hz
Waveform Power spectrum Autocorrelation
Pitch frequency Pitch period
F0 : 160 Hz
Lag (ms)Frequency (Hz)TIme (ms)150
WEAKPITCHES
STRONGPITCHES
Burst length, interburst interval
Higher-order interval pattern
Multiplexed intervals
Interspike interval code
PST or latency pattern
Spike latency
reference times
Interneural synchrony
Average discharge rateRate-
channelcodes
Neural pulse codes
Temporalpatterncodes
Time-of-arrival
codes
Codes are defined in termsof their functional roles
What constitutes a difference that makes a difference?
What spike train messages have the same meanings?(functional equivalence classes)
Temporal codes are neuralcodes in which timings ofspikes relative to each otherare essential to theirinterpretation.
global workspace
highersemantic
resonances
perceptualresonances
sensoryinputs
motoroutputs
ENVIRONMENT
controlledvariables
(consequencesof actions)
uncontrolledvariables(contingent events)
motorpreparations
pragmaticevaluations
deliberationplanning
long-termmemory
Primary sensorypathways
Unimodaland multimodalassociationareas
Motor systems
memory
memory
memory
memory
Limbic & paralimbic areas
Frontal cortical areas
Neural resonances
40
50
40
10
0 1000 2000 3000 4000 5000
20
1000
2000
1000
200
0 100 200 300 400 5000
50
All-order interval (ms)Peristimulus time (ms)
PST Histograms Interval Histograms
4 Hz
8 Hz
16 Hz
32 Hz
64 Hz
Phase-locking of an LGN unit to adrifting sinusoidal gratingTemporal
modulationfrequency
Raw spike train data courtesy of Andrzej Przybyszewski & Dan Pollen
• Depending upon the self-modification process, adaptive systems change in different ways.
• They become tuned to their environments,on the percepton the action sideinternally: anticipating events, forecasting effects
• New sensors create new linkages with the external world new perceptual primitivesnew observablesnew modes of adjustment
• New effectors create new modes of action
Sensing ~ measurement
Adaptation ~ adjustmentAdaptive systems
A B
Resonantstate | A
Resonantstate | B
Contingentsensory input
Epistemiccut
S2
S3
S1
R2
R3
R1
CNS
Potentialstimuli
Potentialresponses
A BStimulus-driven switching betweenreverberant circuits (after Hebb, 1965)
Reverberant patternsas switchable eigenstates
Motorresponse
for A
Motorresponse
for B
Switching between reverberant states
receptor surfaces
Evaluation in termsof basic system-goals
(limbic system)
Evaluation in termsof manifold implications
(associations, planscognitive schemas)
Buildup ofsensoryimages
Sensorytransduction
Structure ofenvironmental
events
Attentionalfacilitationof imageformation
Attentionalfacilitationof imageformation
self-sustainingpatterns
Earlysensorycodng
Functional organization of the perceptual side
Frequency ranges of (tonal) musical instruments
27 Hz 4 kHz262Hz
440Hz
880Hz
110Hz
10k86543210.50.25
Measurement and tuning
Measurementmediates interactions with external world
permitting behavior contingent upon
perception
Tuninginvolves adjustment of internal relations to
external relations, i.e. adaptive resonance
Adaptive systems that create their own measurementsare possible (we may be such systems)
It is possible to envision brains and minds as resonant systems that operate on patternsrather than coupled via energetic relations
Areas of self-modifying media
Self-modifying computersCoevolution between humans and computersEmergent human-machine couplingsPask’s Conversation theoryComputers need means of independently
accessing the world and creating theirown concepts (epistemic autonomy)
Self-organizing materialsElectrochemicalFerromagneticBiological-silicon interfacesIntelligent materials
Mixed digital-analog feedback systems
40
50
40
10
0 1000 2000 3000 4000 5000
20
1000
2000
1000
200
0 100 200 300 400 5000
50
All-order interval (ms)Peristimulus time (ms)
PST Histograms Interval Histograms
4 Hz
8 Hz
16 Hz
32 Hz
64 Hz
Phase-locking of an LGN unit to adrifting sinusoidal gratingTemporal
modulationfrequency
Raw spike train data courtesy of Andrzej Przybyszewski & Dan Pollen
Est. Error: 1.9%Est. Error: 0.4%
1000 50 10050
Interspike interval (ms) Interspike interval (ms)
16 Hz5743 intervals
32 Hz467 intervals
Temporal period estimation for an LGN unit(sinusoidal luminance modulation)
20
40
60Rate tuning curve
Frequency (Hz)10 10010.1
Rate(spikes/s)
Raw spike train data courtesy of Andrzej Przybyszewski & Dan Pollen
Phase-locking in visual thalamus (LGN)
Stimulus:Drifting sinusoidalgratings
Phase-locking in visual neurons(Horseshoe crab ommatidium)
Miller, Ratliff, and Hartline (1961) How cells receive stimuli.Scientific American 215(3):222-238.
Neural timing nets
FEED-FORWARD TIMING NETS• Temporal sieves• Extract (embedded) similarities• Multiply autocorrelations
RECURRENT TIMING NETS• Build up pattern invariances• Detect periodic patterns• Separate auditory objects
Relative delay τ
Timet
Sj(t)
Si(t)
Si(t) Sj(t - τ)
individualmultiplicative
term
Si(tm) Sj(tm - t)Στ
convolutiontime-series
term m
two setsof input
spike trains
In p u t t im e s e q u e n c e
A l l t im e d e la y s p r e s e n t
τ0
τ1
τ2
τ3
C o in c id en c e
u n its
D irec t in p u ts
R ec u r ren t ,
in d irec t in p u ts
T im e p attern s rev erb erate
th ro u g h d elay lo o p s
Vowel [er]F0 = 125 HzPeriod = 8 ms
Period = 10 ms
Vowel [ae]F0 = 100 Hz
Build-up and separation of two auditory objects
Two vowels with different fundamental frequencies (F0s) are added together and passed through the simple recurrent timing net. The two patterns build upIn the delay loops that have recurrence times that correspond to their periods.
.
Time (ms)
0 10 20 30 40 50Time (msec)
Methodological issues:What distinguishes sensing from other kinds
of informational operations?
A sensing process must be contingent, it must have two or more possible outcomes to reduce uncertainty, whereas
A computation (formal operation) must be logically-determined, it must always produce the same outcome given the same initial state
Sensing vs computingContingent vs. logically-necessary “truths”
•Digital computers presently are capable of recombination-based creativity, but do not presently create new primitives for themselves.
• Brains, on the other hand, are self-modifying systems with rich analog dynamics that can serve as substrates for formation of new informational primitives.
• Contemplation of self-modifying systems is essential if we are to construct artificial systems that can create meaning for themselves.
• We need such systems when problems are ill-defined, or when we desire open-ended creative possibilities.
Computers and brains
global workspace
highersemantic
resonances
perceptualresonances
sensoryinputs
motoroutputs
ENVIRONMENT
controlledvariables
(consequencesof actions)
uncontrolledvariables(contingent events)
motorpreparations
pragmaticevaluations
deliberationplanning
long-termmemory
Primary sensorypathways
Unimodaland multimodalassociationareas
Motor systems
memory
memory
memory
memory
Limbic & paralimbic areas
Frontal cortical areas
Neural resonances
• We discuss the semiotics and functional organization of different adaptive systems.
• Adaptive systems reorganize their internal structure in order to improve their performance.
• We consider how systems with sensors, effectors, and coordinative faculties can adaptively modify their internal structures and functions.
• We consider how this adaptivity leads to emergent functions and behaviors.
Overview I: Measurement in adaptive systems
• Creativity has two levels: 1) Recombination of existing primitives
2) De novo creation of new kinds of primitives
• Inherent tradeoffs:
Specifiability vs. autonomy
Predictability/reliability vs. creativity
Overview IV: Creativity, autonomy, and specification
Homeostat
Grey Walter's device
Conceptions of “emergence”
• Structural emergence (appearance of new structures, org. levels)
• Computational emergence (unexpected results)
• Thermodynamic emergence (dissipative systems)
• Functional emergence (flight, color vision)
• Emergence-relative-to-a-model (perspectivist, operationalist)
• Appearance of new structures, functions, behaviors
• Novelty that was not predictable from what came before
Varieties
Methodological issues
• How can we identify the existence of information processing operations in artificial and natural systems?
•How can we distinguish measurement, computation, and effector operations from each other in an unknown material system?
•How can we detect changes in these functionalities, such that we know that our devices or organisms have modified them adaptively?
•We need operational distinctions.•We need to be able to parse a state-transition graph.
Recognizing determinate & contingent events
State-transitionsand
observer-operations
How do we distinguishmeasurementsand computations(such that wecan alsodetect changesin system behavior)?
referencestate
R1
A
B
two or morepossible
outcomes
Measurement Computation Prediction
R2
C
D
PA =
Test
Epistemic cuts(points of contingency)
PB
observed"pointerreading"
B
Emergence relative to an observer:What does the observer have to do to his/her own model to continue successfully predicting the material system’s behavior?
ENVMODELobservablessensorsPredictability Deviation
Recovery ofpredictability
new sensor evolves in devicestability stability
Observer
Organismor
Device
Evolution of observer's modelV2V1Original state spaceAdd new states to V2(no new state variables)V1V2V3Addition ofnew statevariable V3semantic
emergence
V2V1larger Nwithinsame D
larger Nwithinlarger D
syntacticemergence
same Nwithinsame D,differentstate-transitions
N = number of statesD = dimensionality of state space
Opening up the sensory interface:
Break-out strategies for creating new observables
1) construction of new sensors
2) modification of existing sensors
3) interposition of sensory prostheses
4) active measurements
5) creation of new internal sensors
coordinationfeature vector
performance
adaptiveconstruction of a
sensoryprosthesis
∆existingsensors
action vector
environment
test
structuraldevice
boundary
functionaldevice
boundary
prosthesis
Prosthesis:augmentation of functionalities
All technology is prosthesis.
Operational states and procedures in a scientific model
sequence of statesfollowing initial state A
sequence of states after B
referencestateR
1
A
Btwo or morepossible outcomes("pointer readings")
MeasurementComputation
Prediction
R2
C
D
PA =
Test
Epistemic cuts(points of contingency)
PB
Explicate realm of symbols (well-defined)
Implicate realm of material process
(ill-defined)
R
sequence of computationswith initial state A
sequence of computationswith initial state B
referencestate A
B
actions based on theoutcome of the computations
(contingent upon A)
actions based on theoutcome of the computations
(contingent upon B)two or more
possible outcomes("pointer readings")
R1
R2
R3
different preparedreference states
result in differentmeasurements
intentions to make measurementsR, R1, R2, or R3 (motor commands)
physical actions taken to bring thesystem into reference statesR1, R2 or R3 (motor actions)
(R is the passive reference state,without any active preparation)
R1
R2
R3
Computation
"preparingthe system"
Measurement Action
Active measurement
coordinationdiscretefeatures
test
performance
adaptiveconstruction ofnew sensors
∆internalsensors
action vector
environment
structural boundaryof device
internaliconic analog
representations
analog-digitalboundary
Neural assemblies as internal sensors