Project

Back to overview

Inference and Learning with Spiking Neurons

English title Inference and Learning with Spiking Neurons
Applicant Pfister Jean-Pascal
Number 150637
Funding scheme SNSF Professorships
Research institution Institut für Physiologie Medizinische Fakultät Universität Bern
Institution of higher education University of Berne - BE
Main discipline Biophysics
Start/End 01.09.2014 - 31.08.2018
Approved amount 1'437'698.00
Show all

All Disciplines (2)

Discipline
Biophysics
Neurophysiology and Brain Research

Keywords (10)

learning; normative approach; Bayesian inference; long-term plasticity; spiking neuron model; Astrocytes; short-term plasticity; graphical model; synaptic plasticity; recurrent network

Lay Summary (French)

Lead
Le cerveau reçoit en continu une telle quantité d'information qu'il lui est nécessaire d'extraire les informations importantes et de les représenter de façon adéquate. Ce défi d'extraction d'information est d'autant plus sérieux lorsqu'on réalise que les informations sensorielles sont souvent bruitées ou ambiguës. Plusieurs travaux de recherche ont montré que le cerveau est capable de traiter l'incertitude liées aux stimuli de façon optimale. Cependant, nous ne savons pas comment l'incertitude est représentée dans le cerveau et comment le processus d'extraction d'information (appelée inférence probabiliste) est implémenté au niveau des neurones et des synapses.
Lay summary
Contenu et objectifs du travail de recherche

Le but de ce project est de mieux comprendre comment l'incertitude est représentée dans le cerveau et comment l'inférence probabiliste est implémentée au niveau des neurones et des synapses. Pour s'attaquer à ce problème, nous développerons de nouveaux modèles mathématiques de neurones et de synapses.  D'une part, ces modèles seront biologiquement plausibles puisque nous les validerons avec des données in vitro et in vivo. D'autre part, ces modèles seront pertinents au niveau computationel puisque des réseaux de tels neurones seront capables d'effectuer une inférence probabiliste.

Contexte scientifique et social du projet de recherche

Ce travail permettra de mieux comprendre les principes fondamentaux qui gouvernent la dynamique neuronale et la plasticité synaptique. Cette étude apportera une nouvelle lumière sur la façon dont les neurones biologiques effectuent une inférence probabiliste. Ce projet a une double pertinence. Du côté biologique, ce travail permettra d'effectuer des prédictions expérimentales qui sont testables au niveau electrophysiologique. Au niveau computationel, ce projet développera de nouveaux algorithmes d'inférence, ouvrant ainsi la porte au développement de nouveaux logiciels de reconnaissance automatique.
Direct link to Lay Summary Last update: 17.07.2014

Lay Summary (English)

Lead
One of the biggest challenge faced by the brain is to make sense of the perceived environment and extract relevant features from an ambiguous environment. For example, reconstructing the 3D shape of an object which is only perceived on a 2D retina or extracting the melody of a single instrument when many of them are playing simultaneously are intrinsically ambiguous tasks. There are several lines of evidence showing that at the level of sensory processing as well the level of motor control or even at the level of cognitive reasoning, the brain deals with uncertainty in an optimal way. It is however unclear how uncertainty is represented in the brain and networks of neuron manage to perform probabilistic inference (i.e. feature extraction) from noisy sensory inputs.
Lay summary
Aim of the project

The goal of this project is to better understand how uncertainty is represented in the brain and how probabilistic inference and learning is implemented at the level of spiking neurons and synapses. In order to tackle this question, we will develop new mathematical models of neurons and synapses which are at the same time biologically plausible (we will validate those models with in vitro and in vivo data) as well as  computationally relevant (i.e. the neural network can perform probabilistic inference).

Scientific and social context of the research project


This work will provide new insights on the key principles that govern neuronal dynamics and synaptic plasticity. In particular, this study will shed a new light on how biological neurons can perform probabilistic inference. This project has a double relevance. On the biological side, it will produce predictions that can be tested electrophysiologically. On the computational side, it will develop new inference algorithms that can be implemented in other devices (such as silicon chips) and have potentially a wide range of applications. 
Direct link to Lay Summary Last update: 17.07.2014

Responsible applicant and co-applicants

Employees

Publications

Publication
Nonlinear Bayesian filtering and learning: a neuronal dynamics for perception
Kutschireiter Anna, Surace Simone Carlo, Sprekeler Henning, Pfister Jean-Pascal (2017), Nonlinear Bayesian filtering and learning: a neuronal dynamics for perception, in Scientific Reports, 7(8722), 1-13.
Spike-Timing Dependent Plasticity, Learning Rules.
Senn Walter, Pfister Jean-Pascal (2015), Spike-Timing Dependent Plasticity, Learning Rules., in Jaeger Dieter, Jung Ranu (ed.), Springer, New York, 2825-2832.
A Neural Implementation for Nonlinear Filtering
Kutschireiter Anna, Surace Simone Carlo, Sprekeler Henning, Pfister Jean-Pascal (2015), A Neural Implementation for Nonlinear Filtering, in ArXiv, (arXiv:1508), 1.
A Statistical Model for In Vivo Neuronal Dynamics
Surace Simone Carlo, Pfister Jean-Pascal (2015), A Statistical Model for In Vivo Neuronal Dynamics, in Plos One, 10(11), 1-21.
Reinforcement Learning in Cortical Networks
Senn Walter, Pfister Jean-Pascal (2015), Reinforcement Learning in Cortical Networks, Springer, New York.
How to avoid the curse of dimensionality: scalability of particle filters with and without importance weights
Surace Simone Carlo, Kutschireiter Anna, Pfister Jean-Pascal, How to avoid the curse of dimensionality: scalability of particle filters with and without importance weights, in SIAM Review, 1-13.
Online Maximum Likelihood Estimation of the Parameters of Partially Observed Diffusion Processes
SuraceSimone Carlo, PfisterJean-Pascal, Online Maximum Likelihood Estimation of the Parameters of Partially Observed Diffusion Processes, in IEEE Transactions on Automatic Control, 1-15.

Collaboration

Group / person Country
Types of collaboration
Birdsong Group, Institute of Neuroinformatics, University of Zurich / ETH Zurich Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Ciocchi group, Department of Physiology, Bern Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
Mante Group, Institute of Neuroinformatics, University of Zurich / ETH Zurich, Switzerland Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Gosh Group, Institute of Neuroinformatics, UZH Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Computational Neuroscience Group, Universitat Pompeu Fabra Spain (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Psycholinguistics Laboratory, UZH Switzerland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
TU Berlin Germany (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication
Cortical Processing Laboratory, UCL Great Britain and Northern Ireland (Europe)
- in-depth/constructive exchanges on approaches, methods or results
- Publication

Scientific events

Active participation

Title Type of contribution Title of article or contribution Date Place Persons involved
Cosyne 2018 Poster STDP for stochastic synapses: an empirical Bayes approach 01.03.2018 Denver, United States of America Jegminat Jannes;
Cosyne 2018 Poster Nonlinear filtering and learning for point emission processes 01.03.2018 Denver, United States of America Kutschireiter Anna;
Bernstein conference Poster Nonlinear filtering and learning for point emission processes 12.09.2017 Goettingen, Germany Surace Simone Carlo; Pfister Jean-Pascal; Kutschireiter Anna;
Computational Neuroscience meeting CNS*2017 Poster Optimal refractoriness from a rate-distortion perspective 15.07.2017 Antwerp, Belgium Pfister Jean-Pascal; Surace Simone Carlo;
International Conference on Mathematical Neuroscience Poster The Neural Particle Filter 30.05.2017 Boulder, United States of America Surace Simone Carlo; Kutschireiter Anna; Pfister Jean-Pascal;
Cosyne 2018 Poster Spike adaptation as the optimal neural code 01.03.2017 Denver, United States of America Shen Hui-An; Surace Simone Carlo;
Cosyne 2017 Poster Bayesian Spike-Timing Dependent Plasticity 23.02.2017 Salt Lake City, United States of America Pfister Jean-Pascal; Jegminat Jannes;
ZNZ Symposium Poster Unweighted particle filtering: biological relevance and computational challenges 15.09.2016 Zurich, Switzerland Kutschireiter Anna; Pfister Jean-Pascal;
ZNZ Symposium Poster Spike-timing and Property Dependent Plasticity 15.09.2016 Zurich, Switzerland Pfister Jean-Pascal; Dziennik Peter;
International Conference on Mathematical Neuroscience Poster Approximate sampling-based Bayesian inference in a recurrent neuronal network 30.05.2016 Antibes, France Kutschireiter Anna; Pfister Jean-Pascal;
International Conference on Mathematical Neuroscience Talk given at a conference Suppression in a normative model of spike-timing dependent plasticity 30.05.2016 Antibes, France Dziennik Peter; Pfister Jean-Pascal;
Computational Neuroscience meeting CNS*2015 Poster Approximate nonlinear filtering with a recurrent neural network 18.07.2015 Prague, Czech Republic Kutschireiter Anna;
Cosyne 2015 conference Poster A Flexible and Tractable Statistical Model for In Vivo Neuronal Dynamics 05.03.2015 Salt Lake City, United States of America Pfister Jean-Pascal;


Self-organised

Title Date Place

Awards

Title Year
Best poster award for the CNS conference 2015

Use-inspired outputs


Start-ups

Name Year

Associated projects

Number Title Start Funding scheme
179060 Filtering with Spiking Neurons 01.09.2018 SNSF Professorships
137200 Normative theory of synaptic plasticity across multiple time scales 01.10.2011 Ambizione
175644 Bayesian synapses 01.10.2018 Project funding (Div. I-III)

Abstract

One of the biggest challenge faced by the brain is to make sense of the perceived environment. Even though this is done in a seemingly effortless way, the required computation steps to make sense of the environment and extract relevant features are far from being trivial. How to reconstruct the 3D shape of an object which is only perceived on a 2D retina? How to recognize an object if only part of it is observed or if it is observed from a new perspective? How to optimally combine multi-sensory cues given that each sensor has its own reliability? How to extract the melody of a single instrument when many of them are playing simultaneously? Interestingly, all those psychophysical tasks which deal with uncertainty can be formulated in a generic probabilistic framework in which the sensory observations are assumed to be generated by some unobserved (hidden) causes. The task in this generative model perspective is to invert the model by inferring the hidden features (causes) given the observations and learn the parameters of the model. This inference procedure can be expressed mathematically with Bayes' rule. Despite the increasing interest in this normative approach, there is still one important question that remains unanswered: how is this learning and inference implemented at the level of single synapses and at the level of spiking neurons? This proposal aims at filling this gap and divides this challenging task into three projects. The first project will consider a generative model where the causes are dynamic and sparsely distributed. This model will generalize existing generative model approaches (such as slow-feature analysis) by assuming that the stationary distribution of each hidden variable is not restricted to a Gaussian distribution. The second project will address the question of inference and learning at the level of single synapse and in the presence of spiking neurons. The last project will combine the computational relevance of the first project and the biological plausibility of the second project thereby deriving a unifying framework for inference and learning with spiking neurons at the network level.
-