Download presentation
Presentation is loading. Please wait.
Published byArnold Burns Modified over 9 years ago
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. Specifically, we will look at the functional and anatomical asymmetries in (extrinsic and intrinsic) connections and their implications for spectral responses. Canonical circuits for predictive coding Karl Friston, University College London
2
Overview The free-energy principle action and perception predictive coding with reflexes The anatomy of inference graphical models canonical microcircuits Functional asymmetries extrinsic connections intrinsic connections
3
“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 Bayes Geoffrey Hinton Richard Feynman From the Helmholtz machine to the Bayesian brain and self-organization Richard Gregory Hermann von Helmholtz Ross Ashby
4
Minimizing prediction error Change sensations sensations – predictions Prediction error Change predictions Action Perception
5
Prior distribution Posterior distribution Likelihood distribution temperature Prediction errors – the ‘Bayesian thermostat’ 20406080100120 Perception Action
6
Overview The free-energy principle action and perception predictive coding with reflexes The anatomy of inference graphical models canonical microcircuits Functional asymmetries extrinsic connections intrinsic connections
7
A simple hierarchy Generative models whatwhere Sensory fluctuations
8
Generative model Model inversion (inference) A simple hierarchy Descending predictions Descending predictions Ascending prediction errors From models to perception Expectations: Predictions: Prediction errors: Predictive coding
9
Haeusler and Maass: Cereb. Cortex 2006;17:149-162Bastos et al: Neuron 2012; 76:695-711 Canonical microcircuits for predictive coding
10
frontal eye fields geniculate visual cortex retinal input pons oculomotor signals Errors (superficial pyramidal cells) Expectations (deep pyramidal cells) Top-down or backward predictions Bottom-up or forward prediction error proprioceptive input reflex arc Perception David Mumford Predictive coding with reflexes Action
11
Overview The free-energy principle action and perception predictive coding with reflexes The anatomy of inference graphical models canonical microcircuits Functional asymmetries extrinsic connections intrinsic connections
12
superficial deep Errors (superficial pyramidal cells) Expectations (deep pyramidal cells) 020406080100120 0 0.05 0.1 0.15 0.2 0.25 0.3 020406080100120 0 1 2 frequency (Hz) 020406080100 0 2 4 6 8 10 12 14 spectral power Forward transfer function 020406080100 0 1 2 3 4 5 6 frequency (Hz) spectral power Backward transfer function Andre Bastos V4V1
13
Errors (superficial pyramidal cells) Expectations (deep pyramidal cells) Linear or driving connections Nonlinear or modulatory connections superficial deep NMDA receptor density
14
020406080100120 0 1 2 frequency (Hz) Errors (superficial pyramidal cells) Expectations (deep pyramidal cells) superficial deep Nonlinear (cross frequency) coupling 020406080100120 0 1 2 frequency (Hz) 020406080100120 0 0.05 0.1 0.15 0.2 0.25 0.3 γ
15
STN M1 STN M1 γ γ On dopamineOff dopamine M1 STN M1 STN Bernadette Van Wijk
16
Summary Hierarchical predictive coding is a neurobiological plausible scheme that the brain might use for (approximate) Bayesian inference about the causes of sensations Predictive coding requires the dual encoding of expectations and errors, with reciprocal (neuronal) message passing Much of the known neuroanatomy and neurophysiology of cortical architectures is consistent with the requisite message passing In particular, the functional asymmetries and laminar specificity of intrinsic and extrinsic connections provide a formal perspective on spectral asymmetries and cross frequency coupling in the brain.
17
Thank you And thanks to collaborators: Rick Adams Andre Bastos Sven Bestmann Harriet Brown CC Chen Pascal Fries Lee Harrison Stefan Kiebel James Kilner Andre Marreiros Jérémie Mattout Rosalyn Moran Will Penny Klaas Stephan Bernadette Van Wijk And many others
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.