PhD Program: course in Computational Neuroscience

Context

Computational neuroscience is an expending field that is proving to be essential in neurosciences. The aim of this course will be to provide a common solid background in computational neurosciences. The course will comprise historical recall of the field and a description of the different modelling approaches that are currently developed, including details about their specificities, limits and advantages.

Objective

The course aims at introducing students with the major tools that will be necessary during their thesis to model or analyze their neuroscientific results. While it will start by a short, generic introduction, we will then explore different systems at different scales. On the first day, we will study the different possible regimes in which a single neuron can behave, while progressively introducing the theory of dynamical systems to understand these more globally. Then, during the second day, we will introduce methods to analyze neuroscientific data in general, such as Bayesian methods and information theory. This will be implemented by simple practical examples.

Language of intervention

English

Number of hours

~20 hours (session 1=7 + session 2=7 + session 3=4)

Max participants

15 for the practical sessions (afternoon Day 2 and Day 3), unlimited for theoretical courses

Public priority

PhD students

Public concerned

PhD students, interested M2 students and postdocs

Location

Institut des Neurosciences de la Timone (INT)

Keywords

neuronal modelling, neural circuit modelling, information theory, decoding and encoding

Targets

Understanding how computational modelling can be used to formulate and solve neuroscience problems at different spatial and temporal scales; learning the formal notions of information, encoding and decoding and experimenting their use on toy datasets

Program

First session: Introduction to modeling single neurons (morning); An introduction to neural masses: modeling assemblies of neurons up to capturing collective oscillations and resting state dynamics in a mean-field model - presentation of the Virtual Brain software (afternoon) - Second session: An overview on “What is encoding?” “What is decoding?”: formalization of the notion of information in neural activity; shared and transferred information; integration, segregation and complexity (morning). Bayesian probabilities, the Free-energy principle and Active Inference, with practical demonstrations in python (afternoon). Third session: the problem of information estimation in practice. Practical exercices in Matlab: estimating entropy and stimulus decodability from spike trains; comparing coding hypotheses (morning).

Pre-required

Basic knowledge of statistics and probability and calculus (differential equations,…) is useful, but steps will be explained and complex math avoided as much as possible. Practical exercises are in python and/or MATLAB, so basic knowledge of these environments is a plus.

program

day 1 : 2018-03-26 : an introduction to Computational Neuroscience

• 09:30-12:30 = Introduction to modeling single neurons (LaP)

• 14:00-17:00 = An introduction to neural masses: modeling assemblies of neurons up to capturing resting state dynamics in a mean-field model - presentation of the Virtual Brain software (DaB)

day 2 : 2018-03-27 : Information theory / bayesian models

• 09:15-10:30 = An overview on “What is encoding?” “What is decoding?”: formalization of the notion of information in neural activity (DaB)

• 11:00-12:15 = (…continued after the coffee break: ) Live information! From sharing information to transferring information (and a glimpse into the zoo of higher-order friends) (DaB)

• 14:00-17:10 = Probabilities, the Free-energy principle and Active Inference (LuP).

day 3 : 2018-03-28 : Practical course on Information theory

• 09:30-12:30 = Practical course on Information theory (DaB)

More material related to the course

day 1 - morning : the single neuron

• site du livre de Gerstner et al “Neuronal Dynamics”: http://neuronaldynamics.epfl.ch/

• A (longer) introduction to the Hodgkin-Huxley model in three steps by Dr Stefano Luccioli

  • http://neuro.fi.isc.cnr.it/uploads/TALKS/lez1.pdf

  • http://neuro.fi.isc.cnr.it/uploads/TALKS/lez2.pdf

  • http://neuro.fi.isc.cnr.it/uploads/TALKS/lez3.pdf

• An interactive course with Wulfram Gerstner https://www.edx.org/course/neuronal-dynamics-computational-epflx-bio465-1x

• His book ONLINE http://cn.epfl.ch/~gerstner/NeuronalDynamics-MOOC1.html

day 1 - afternoon : neural mass models

• Another interactive course @ Washington University https://www.coursera.org/course/compneuro

• Collection of didactic material for the EU FP7 ITN Neural Engineering Transformative Technology http://www.neural-engineering.eu/training/index.html

• Didactic material from Lab in Computational Neuroscience http://neuro.fi.isc.cnr.it/index.php?page=didactic-material

• A open source simulator of a whole brain which runs on your laptop, “The Virtual Brain”: http://thevirtualbrain.org

day 2 - morning : information theory

• The best book on information theory and decoding, freely available directly from the author: http://www.inference.phy.cam.ac.uk/itprnn/book.html

• a gentle introduction to bayesian methods : https://homepages.inf.ed.ac.uk/pseries/Peg_files/Chapter9_SotiropoulosSeries.pdf

day 2 - afternoon : bayesian models

• an interesting read : http://cognitrn.psych.indiana.edu/busey/q551/PDFs/PredictivCodingRaoBallard.pdf

• a tutorial on free-energy : some exercises : http://www.sciencedirect.com/science/article/pii/S0022249615000759

• solutions to the tutorial : https://laurentperrinet.github.io/sciblog/posts/2017-01-15-bogacz-2017-a-tutorial-on-free-energy.html

contacts

• LaP: Laurent Pezard «Laurent.Pezard@univ-amu.fr»

• DaB: Demian Battaglia «demian.battaglia@univ-amu.fr», INS

• LuP: Laurent Udo Perrinet «laurent.perrinet@univ-amu.fr», INT

• PhD program: Nicole Malfait «Nicole.Malfait@univ-amu.fr», Anna Montagnini «anna.montagnini@univ-amu.fr»