hierarchal brain architectures and the bayesian brain karl friston, university college london
DESCRIPTION
Hierarchal brain architectures and the Bayesian brain Karl Friston, University College London. - PowerPoint PPT PresentationTRANSCRIPT
Slide 1
How much about our interactions with and experience of our world can be deduced from basic principles? This talk reviews recent attempts to understand the self-organised behaviour of embodied agents, like ourselves, as satisfying basic imperatives for sustained exchanges with the environment. In brief, one simple driving force appears to explain many aspects of action and perception. This driving force is the minimisation of surprise or prediction error that in the context of perception corresponds to Bayes-optimal predictive coding (that suppresses exteroceptive prediction errors) and in the context of action reduces to classical motor reflexes (that suppress proprioceptive prediction errors). We will look at some of the implications for the anatomy of this active inference, in terms of large-scale anatomical graphs and canonical microcircuits, and then turn to some examples of active inference such as perceptual categorisation, action and its perception. Hierarchal brain architectures and the Bayesian brain
Karl Friston, University College London
Overview
The free-energy principleaction and perceptionpredictive coding with reflexes
The anatomy of inferencegraphical modelscanonical microcircuits
Some examplesperceptual categorizationomission responsesaction observation
Objects are always imagined as being present in the field of vision as would have to be there in order to produce the same impression on the nervous mechanism - von Helmholtz
Thomas BayesGeoffrey HintonRichard FeynmanFrom the Helmholtz machine to the Bayesian brain and self-organizationRichard Gregory
Hermann von Helmholtz Ross Ashby
Minimizing prediction error
Change sensationssensations predictionsPrediction errorChange predictionsActionPerception
Prior distributionPosterior distributionLikelihood distributiontemperature
Action as inference the Bayesian thermostat
20406080100120
Perception
Action
Overview
The free-energy principleaction and perceptionpredictive coding with reflexes
The anatomy of inferencegraphical modelscanonical microcircuits
Some examplesperceptual categorizationomission responsesaction observation
A simple hierarchyGenerative models
whatwhereSensory fluctuations
Generative modelModel inversion (inference)A simple hierarchyExpectations:Predictions:Prediction errors:
DescendingpredictionsAscending prediction errorsFrom models to perception
Haeusler and Maass: Cereb. Cortex 2006;17:149-162Bastos et al: Neuron 2012; 76:695-711Canonical microcircuits for predictive coding
frontal eye fieldsgeniculatevisual cortexretinal inputponsoculomotor signals
Errors (superficial pyramidal cells)Expectations (deep pyramidal cells)Top-down or backward predictionsBottom-up or forward prediction errorproprioceptive inputreflex arcPerception
David MumfordPredictive coding with reflexesAction
superficialdeep
Errors (superficial pyramidal cells)Expectations (deep pyramidal cells)
02040608010012000.050.10.150.20.250.3020406080100120012x 10-4frequency (Hz)
02040608010000.511.522.53spectral powerV1 Autospectra (SPC)02040608010002468101214spectral powerForward transfer function02040608010005101520frequency (Hz)spectral powerV4 Autospectra (DPC)0204060801000123456frequency (Hz)spectral powerBackward transfer function
Errors (superficial pyramidal cells)Expectations (deep pyramidal cells)
Linear or driving connectionsNonlinear or modulatory connectionssuperficialdeep
NMDA receptor density
Biological agents resist the second law of thermodynamics
They must minimize their average surprise (entropy)
They minimize surprise by suppressing prediction error (free-energy)
Prediction error can be reduced by changing predictions (perception)
Prediction error can be reduced by changing sensations (action)
Perception entails recurrent message passing in the brain to optimize predictions
Action makes predictions come true (and minimizes surprise)Overview
The free-energy principleaction and perceptionpredictive coding with reflexes
The anatomy of inferencegraphical modelscanonical microcircuits
Some examplesperceptual categorizationomission responsesaction observation
Generating bird songs with attractorsSyrinxHVCtime (sec)FrequencySonogram0.511.5
Hidden causesHidden states
102030405060-505101520prediction and error102030405060-505101520hidden statesBackward predictionsForward prediction error102030405060-10-505101520causal statesPredictive coding
stimulus0.20.40.60.82000250030003500400045005000time (seconds)
Perceptual categorization
Frequency (Hz)Song a
time (seconds)Song b
Song c
Perceptual inference and sequences of sequencesSyrinxNeuronal hierarchy
Time (sec)Frequency (KHz)sonogram0.511.5
Prediction errorSensory statesHidden states
omission and violation of predictionsStimulus but no perceptPercept but no stimulusFrequency (Hz)stimulus (sonogram)25003000350040004500Time (sec)Frequency (Hz)percept0.511.525003000350040004500500100015002000-100-50050100peristimulus time (ms)LFP (micro-volts)ERP (prediction error)without last syllableTime (sec)percept0.511.5500100015002000-100-50050100peristimulus time (ms)LFP (micro-volts)with omission
Overview
The free-energy principleaction and perceptionpredictive coding with reflexes
The anatomy of inferencegraphical modelscanonical microcircuits
Some examplesperceptual categorizationomission responsesaction observation
Prior distributiontemperatureAction as inference the Bayesian thermostat
20406080100120
Perception:
Action:
visual inputproprioceptive inputAction with point attractors
Descendingproprioceptive predictionsExteroceptive predictions
00.20.40.60.811.21.40.40.60.811.21.4actionposition (x)position (y)00.20.40.60.811.21.4observationposition (x)Heteroclinic cycle (central pattern generator)
Descendingproprioceptive predictions
Each movement we make by which we alter the appearance of objects should be thought of as an experiment designed to test whether we have understood correctly the invariant relations of the phenomena before us, that is, their existence in definite spatial relations.
'The Facts of Perception' (1878) in The Selected Writings of Hermann von Helmholtz,Ed.R. Karl, Middletown: Wesleyan University Press, 1971 p. 384
Hermann von Helmholtz Thank you
And thanks to collaborators:
Rick AdamsAndre BastosSven BestmannHarriet BrownJean DaunizeauMark EdwardsXiaosi GuLee HarrisonStefan KiebelJames KilnerJrmie MattoutRosalyn MoranWill PennyLisa Quattrocki Knight Klaas Stephan
And colleagues:
Andy ClarkPeter DayanJrn DiedrichsenPaul FletcherPascal FriesGeoffrey HintonJames HopkinsJakob HohwyHenry KennedyPaul VerschureFlorentin Wrgtter
And many others
Overview
The free-energy principleaction and perceptionpredictive coding with reflexes
The anatomy of inferencegraphical modelscanonical microcircuits
Some examplesperceptual categorizationomission responsesaction observationvisual searches
If percepts are hypotheses, where do we look for evidence?Richard Gregory
saliencevisual inputstimulussamplingSampling the world to minimise uncertaintyPerception as hypothesis testing saccades as experiments
Free energy minimisationminimise uncertainty
Frontal eye fields
Pulvinar salience mapFusiform (what)Superior colliculusVisual cortexoculomotor reflex arc
Parietal (where)
29
Saccadic fixation and salience mapsVisual samplesConditional expectations about hidden (visual) statesAnd corresponding perceptSaccadic eye movementsHidden (oculomotor) states
Perception and Action: The optimisation of neuronal and neuromuscular activity to suppress prediction errors (or free-energy) based on generative models of sensory data.
Learning and attention: The optimisation of synaptic gain and efficacy over seconds to hours, to encode the precisions of prediction errors and causal structure in the sensorium. This entails suppression of free-energy over time.
Neurodevelopment: Model optimisation through activity-dependent pruning and maintenance of neuronal connections that are specified epigenetically
Evolution: Optimisation of the average free-energy (free-fitness) over time and individuals of a given class (e.g., conspecifics) by selective pressure on the epigenetic specification of their generative models.
Time-scaleFree-energy minimisation leading to
Searching to test hypotheses life as an efficient experimentFree energy principleminimise uncertainty
Epilogue
(what we have not covered)
Synaptic gainSynaptic activitySynaptic efficacy
Perception and inferenceLearning and memoryPosterior beliefs and sufficient statistics
Attention and precisionPerception and inference
Learning and memory
Attention and affordance
Sensory attenuation
Random dynamical attractors and ergodic theorem(path integral formulations and principle of least action)Discrete formulations and Markovian processes(optimal decision theory)Continuous formulations and dynamical systems theory(self-organised criticality)The free energy principle
Variational Bayes = ensemble learningGeneralized Bayesian filtering = predictive codingFokker-Planck equation = ensemble dynamics
Sleeping and dreaming(complexity minimisation and synaptic homoeostasis)Interoception and predictive coding(emotional valence and self-awareness)Neuropsychiatry(false inference and failures of sensory attenuation)The free energy principle
Predictive coding and embodied cognition(philosophy)