Home | Lab | Research |Teaching | Publications | Links


Computational Cognitive Neuroscience -- Autumn 2017

Instructor: Peggy Seriès
Lectures :***CHANGE OF ROOM*** : 5.3 Lister weeks 3-7 and 9-11 and Lecture theatre 1 - 7 Bristo Square for week 8. on Mondays and Thursdays in Teviot Lecture Room Medical School doorway 5
Labs : 15.10-17.00 on Tuesdays (5.05 West Lab AT) and Wednesdays (4.12 AT), on wk 2, 4, 6, 8, 10.
On wk1 students will be offered a non compulsory matlab tutorial covering basic matlab prerequisites.
TA & Tutor: Samuel Rupprechter and Frank Karvelis.

This is a 10 points course for MSc level students. There are no prerequisites, no prior knowledge in neuroscience is necessary but some background in statistics, calculus, linear algebra will help, as a well as some knowledge of matlab (or python).

Computational Cognitive Neuroscience is a growing research field. The aim of this course is to learn the tools and concepts that can be used to model the relation between the activity of the brain and cognitive processes. This course will appeal to students who are interested in the basic principles of computation in the human and animal brain, in particular how we can relate the activity of the brain to perception, behaviour or decision-making.

This course differs from / complements NC and NIP in focussing on 'higher level' processes and phenomena (e.g. decision making) and more conceptual models (even if we'll try to stay as close as possible to neurophysiology). It is a good complement to those courses if you are interested in a PhD in computational neuroscience.
The topics discussed in the course are mostly of interest for academic research and although they have inspired machine learning solutions, they are of little direct applicability. Tools from machine learning, though, are commonly used in this field.
Apart from learning about the brain, you will also learn about numeral modelling of differential equations, random processes, decoding techniques, dynamical systems, Bayesian modelling and more.

Resources

  • No proper textbook. We will use Dayan and Abbott's Theoretical Neuroscience, MIT Press and and various journal articles / book chapters that I will provide.

  • If you want to prepare for the course, 2 things would be beneficial:
    i) read a primer about neuroscience, e.g. Brain Facts .
    ii) make sure you can program in Matlab (see below for tutorials). You can use for e.g. : Chris' handout introduction to matlab; or Matlab primer.

Assignments

  • Two assignments (50%) and one report (50%) (no exam).

  • Matlab assignment 1. Deadline: October 27th 4 pm.
  • Related to the Encoding and Decoding lectures.
    The report should look like a scientific report, with description and discussion of the results (not a presentation of the code). The quality of the presentation will be taken into account in the final mark.

  • Matlab assignment 2 distributed: Oct 27th Deadline: November 17th.
    Related to Reinforcement learning lectures.

  • Report: You will have to write a paper based on one or two papers of your choice. The paper(s) should be related to the themes of cognitive neuroscience and have a significant computational component (describe a mathematical model or use simulations). It should describe an original piece of research (not be a review !). Please contact me if you're not sure of your choice or want suggestions. You should explore the context, critically evaluate it, and discuss questions raised by these papers and maybe suggest further experimental or theoretical work.
    Your report should be around 3000 words (4000 words MAX) all included (references, captions etc..). You should write for the interested, but non-specialist reader. You can look at the journal Trends in Neuroscience (TINS) for how to construct such papers. The aim of the paper is that you should demonstrate that you can read a paper in computational cognitive neuroscience, understand its methods, evaluate its claims and place it in perspective.

    Deadline for submitting choice of paper: 8th November 2017. Please send me the pdf of your paper by email.

    Deadline for report: 8th December 2017. please submit both pdf via submit and paper copy to ITO.

    Tips for choosing a paper.
    Examples of 2 good papers from previous years: paper1 (attention), paper2 (hallucinations in migraines).

Labs


The labs will consist mainly of implementation in matlab of simple models of population codes, perception, learning and plasticity and decision making. Attendance and work on this material will help for the assignments. Example answer code for each lab will be distributed after each lab.

Classes

The course slides will be posted here before the class.
  • 18/09/17: Introduction to the course. lecture slides .

  • 21/09/17: Encoding. Overview of the brain, neurons, synapses and electrophysiology. Responses in visual cortex. lecture slides

  • 25/09/17: Encoding: Variability in the brain. Poisson and Gaussian models of spike count distributions. lecture slides

  • 28/09/17: Applications of Encoding. Review of the visual system. introduction to Decoding. Lectures slides. Reading: Visual prostheses for the blind.

  • 01/10/17: Decoding. Application to Brain-Machine interfaces. Basics of Estimation theory. Lectures slides Reading: Reading Minds; more advanced: BMI beyond neuro-prosthetics.

  • 05/10/17: Fisher Information. Applications of the Encoding-Decoding model in research problems. Lectures slides

  • 09/10/17: Models of neurons. Lectures slides.

  • 12/10/17: Models of networks of neurons. Lecture slides.

  • 16/10/17: Supervised learning in networks of neurons. Lecture slides .

  • 19/10/17: Reinforcement learning in computational neuroscience. Lecture slides .

  • 23/10/17: Sam Rupprechter teaching: Reinforcement learning models to understand Depression.

  • 26/10/17: Sam Rupprechter teaching: title TBA

  • 30/10/17: Models of Working Memory.

  • Decision Making.

  • Bayesian Brain Theories and applications.

  • Computational Psychiatry, selected topics
In the meantime, slides from previous year can give a good idea of the content of the future classes:

This page is maintained by Peggy Series.