Rosalyn J. Moran Wellcome Trust Centre for Neuroimaging
1st Workshop on the Free Energy Principle, ION, UCL, July 5th 2012.
Precision in Cortical Message Passing
Outline
Predicting & Estimating Precision under the Free Energy Principle - Laplace and Mean Field Assumptions
Hypothesised Neuronal Implementation & the role of Neuromodulators - Gain effects on primary neurotransmission
Cholinergic Neuromodulation & Certainty Effects on Auditory mismatch negativity - Theoretical simulation of perception
Testing Cholinergic Neuromodulation - DCM characterization of Event Related Responses
Outline
Predicting & Estimating Precision under the Free Energy Principle - Laplace and Mean Field Assumptions
Hypothesised Neuronal Implementation & the role of Neuromodulators - Gain effects on primary neurotransmission
Cholinergic Neuromodulation & Certainty Effects on Auditory mismatch negativity - Theoretical simulation of perception
Testing Cholinergic Neuromodulation - DCM characterization of Event Related Responses
Predicting & Estimating Precision under the Free Energy Principle
Hierarchical, Dynamic & Uncertain causes in the environment generate sensory signals
Different Levels of the hierarchy and/or different sensory signals may confer more precise Information
The Environment
Hierarchical, Dynamic
The Environment
Hierarchical, Dynamic & Uncertain causes generate sensory signals
y y
The InversionEstimate: Hierarchical, Dynamic & Uncertainty of sensory signals to minimise the surprise of the sensory signals
y y
Minimise Free Energy
)),|(||)(()|)(()()|)(()(
)|)(()|(
)|(ln)|()|(
mypqKLmtyptFmtyptF
dtmtypmyH
dymypmypmyH
Minimise SurpriseTime averaged Surprise(Ergodicity)
Minimise F at every point in time
The Brain’s Response to y… A Tractable Problem
States, parameters & noise
Outline
Predicting & Estimating Precision under the Free Energy Principle - Laplace and Mean Field Assumptions
Hypothesised Neuronal Implementation & the role of Neuromodulators - Gain effects on primary neurotransmission
Cholinergic Neuromodulation & Certainty Effects on Auditory mismatch negativity - Theoretical simulation of perception
Testing Cholinergic Neuromodulation - DCM characterization of Event Related Responses
Minimising Free Energy
The Laplace Assumption: The brain assumes gaussian random fluctuations
)),|(||)(()|)(()( mypqKLmtyptF
y
Gradients a function of error terms weighted by the precisions at each level:How might precisions be encoded?
Smooth noise correlations within levels Markov properties between levels
1 5 1015200 25
1 5 1015200 25
1 5 10 152025
Gradients of Free Energy Precision Dependent
y
A multiplicative term that stays within levels:Candidate mechanisms: local lateral inhibition & neuromodulators
Forward prediction error
Backward predictions
Superficial pyramidal cells
Deep pyramidal cells
Perceiving multiple hierarchical levels together: errors can have a greater or lesser effect
Gain control at superficial pyramidal cells y
Neuromodulators: Anatomically deployed to provide input in multiple regionsEg Sarter et al. 2009
Local Glutamate & GABA
Long Range Glutamate
Diffuse projectionsNeuromodulatorsAcetylcholineDopamine
Gain control at superficial pyramidal cells yNeuromodulators: Physiologically equipped to provide gain control
Cholinergic Projectionsfrom BasalForebrain
Dopaminergic Projections from VTA/SNc
Activity at D1 receptorsstimulates adenylyl cyclasemodulating postsynaptic currentsActivity at muscarinic receptors
enhances EPSPs through K-current modulation
Gain control at superficial pyramidal cells yNeuromodulators: Physiologically equipped to provide gain control
Cholinergic Projectionsfrom BasalForebrain
Dopaminergic Projections from VTA/SNc
Presynaptic terminals
Excitatory (AMPA) receptorsModulatory receptorInhibitory (GABAA) receptors
Dendritic spine
errorprecision Precision-weighted error
Outline
Predicting & Estimating Precision under the Free Energy Principle - Laplace and Mean Field Assumptions
Hypothesised Neuronal Implementation & the role of Neuromodulators - Gain effects on primary neurotransmission
Cholinergic Neuromodulation & Certainty Effects on Auditory mismatch negativity - Theoretical simulation of perception
Testing Cholinergic Neuromodulation - DCM characterization of Event Related Responses
Testing error precision modulation by Acetylcholine:The Framework
7 Auditory Stimuli:Pure tones presented in mini-blocks
time
Freq
Mismatch Negativity ~150 ms
Under Placebo & Cholinergic Enhancement
Simulate Experiment
Recognition Dynamics
v1
x1 x2
There was a particular
sound
The sound has dynamics determined
by properties, Frequency and
Amplitude
Sensationsy~
Recognition Dynamics
Testing error precision modulation by Acetylcholine:The Sensory Data
A two level hierarchytimeFreq
v1
x1 x2
Sensationsy~
Testing error precision modulation by Acetylcholine:The Sensory Data
C =4
A two level hierarchytimeFreq
v1
x1 x2
Sensationsy~
Testing error precision modulation by Acetylcholine:The Sensory Data
C = 2
timeFreq
Sensationsy~
Testing error precision modulation by Acetylcholine:The Inversion: assume different precision estimates
Placebo
ACh
timeFreq
Sensations
y~
Testing error precision modulation by Acetylcholine:The Recognition Dynamics under different precision estimates
Placebo
ACh
Time (msec) Time (msec)
Simulated ERP Placebo Simulated ERP ACh
Prec
ision
wei
ghte
d PE
0 50 100 150 200 250 300-80
-60
-40
-20
0
20
40
60
80
0 50 100 150 200 250 300-80
-60
-40
-20
0
20
40
60
80
d1d2d10
Testing error precision modulation by Acetylcholine:The MMN itself under different precision estimates
Time (msec) Time (msec)
Simulated ERP Placebo Simulated ERP ACh
Prec
ision
wei
ghte
d PE
0 50 100 150 200 250 300-80
-60
-40
-20
0
20
40
60
80
0 50 100 150 200 250 300-80
-60
-40
-20
0
20
40
60
80
d1d2d10
-5
0
5
10
15
20
25
Prec
isio
n we
ight
ed P
E
Simulated MMN Placebo Simulated MMN ACh (more Precision)
CertainEnvironmentUntil oddball
More CertainEnvironmentUntil oddball
Tone is predictedTone is predicted
Outline
Predicting & Estimating Precision under the Free Energy Principle - Laplace and Mean Field Assumptions
Hypothesised Neuronal Implementation & the role of Neuromodulators - Gain effects on primary neurotransmission
Cholinergic Neuromodulation & Certainty Effects on Auditory mismatch negativity - Theoretical simulation of perception
Testing Cholinergic Neuromodulation - DCM characterization of Event Related Responses
Testing error precision modulation by Acetylcholine:
7 Auditory Stimuli:Pure tones presented in mini-blocks
time
Freq
Mismatch Negativity ~150 ms
Under Placebo & Cholinergic Enhancement
Real Experiment
Scalp Effects: MMN
Recorded MMN Placebo
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5 *
chan
nel C
21
*
Recorded MMN Galantamine
-5
0
5
10
15
20
25
Prec
isio
n we
ight
ed P
E
Simulated MMN Placebo Simulated MMN Galantamine (more Precision)
CertainEnvironmentUntil oddball
More CertainEnvironmentUntil oddball
Tone is predictedTone is predicted
Physiological & Hierarchical PredictionsRecall:
Forward prediction error
Backward predictions
Superficial pyramidal cells
Deep pyramidal cells
A multiplicative term that stays within levels:Candidate mechanisms: neuromodulators
Forward (Bottom-up) ConnectionBackward (Top-Down) Connection
IFG
A1MTG
IFG
A1 MTG
What layer? What region?
Acetylcholine: Where does it affect network processing?
( )x
( )x
( )v
( )v
Spiny stellate
Deep pyramidal
Superficial pyramidalInhibitory interneuron
Backward connections
Forward connections
Gain Modulation at Supragranular Pyramidal Cells
Gain Modulation at Deep Pyramidal Cells
Acetylcholine: Where does it affect network processing?
Forward (Bottom-up) ConnectionBackward (Top-Down) Connection
IFG
A1MTG
IFG
A1 MTG
What layer? What region?
Forward (Bottom-up) ConnectionBackward (Top-Down) Connection
IFG
A1MTGIFG
A1MTG
simple neuronal model
Slow time scale
fMRIcomplicated neuronal model
Fast time scale
EEG/MEG
),,( uxFdtdx
Neural state equation:
Hemodynamicforward model:neural activityBOLD
Time Domain Data
Electromagneticforward model:
neural activityEEGMEG
LFP
Time Domain ERP Data…
Neural Mass Model
DCM
Acetylcholine: Where does it affect network processing?
Forward (Bottom-up) ConnectionBackward (Top-Down) Connection
IFG
A1MTG
IFG
A1 MTG
What layer? What region?
Forward (Bottom-up) ConnectionBackward (Top-Down) Connection
IFG
A1MTGIFG
A1MTG
( )x
( )x
( )v
( )v
Spiny stellate
Deep pyramidal
Superficial pyramidalInhibitory interneuron
Backward connections
Forward connections
DCM for ERPs : Canonical Microcircuit
Acetylcholine: Bayesian Model Selection
Forward ConnectionBackward Connection
Model 3IFG
MTG A1
IFG
A1
Model 1
MTG
Model 2IF
GM
TG 1A
IFG
A1
MTG
IFG
MTG A1
IFG
A1
MTG
IFG
MTG A1
IFG
A1
MTG
MTG MTG
A1
Model 5 Model 6
IFG
MTG A1
IFG
A1
MTG MTG MTG
A1
IFG
MTG A1
IFG
A1
MTG
Model 7 Model 8IF
GM
TG A1
IFG
A1
MTG
IFG
MTG A1
IFG
A1
MTG
Model 9 Model 10
IFG
MTG A1
IFG
A1
MTG
IFG
MTG A1
IFG
A1
MTG
Model 3 Model 4
Intrinsic Modulation (models 1-6); Extrinsic Modulation (models 7-10)
Rela
tive
Log
Mod
el E
vide
nce
M1 M2 M3 M4 M5 M60
200
400
600
800
1000
∆F = 153
M7 M8 M9 M10
IFG
( )x
( )x
( )v
( )v
Spiny stellate
Deep pyramidal
Superficial pyramidalInhibitory interneuron
Backward connections
Forward connections
Gain Modulation at Supragranular Pyramidal Cells
In A1
Acetylcholine: Direction of Gain Modulation
Placebo
ACh
Mod
ulat
ory
Effe
ct o
f Gal
anta
min
e
Superficial Pyramidal Cell Gain
0.01
0.02
0.03
0.04
0.05
0.06
PlaceboBaseline
Galantamine
*
Summary
Precision estimates enable Bayes optimal perception - Hierarchical inference enables different precision effects at different
levels- Precision estimates control the impact of errors in Free Energy
minimisation under the Laplace Assumption
Neuromodulators are anatomically & physiologically equipped to signal precision in this scheme
Neuromodulatory systems could control precision at different hierarchical levels
Cholinergic Neuromodulation controls gain in superficial pyramidal cells in early sensory regions; conforming to Free Energy Predictions of enhanced precision on sensory prediction errors
Thank You
Karl FristonRay DolanKlaas Enno StephanMkael SymmondsNicholas WrightPablo CampoMethods GroupEmotion Group
Acknowledgments