Systems Computational Neuroscience

Eventtype: Lecture

Frequency: winter semester

Hours per week: 90 min lecture; 60 min homework

Contents

Part 1: Neural reinforcement learning

Part 2: Computational vision

Part 3: Computational motor control

Learning target

Students taking this course will get a grounding in the core ideas and methods of systems computational neuroscience. We will collectively cover key elements of sensory input, using vision as a central example, internal value-based selection and choice, in the form of reinforcement learning, as well as motor control and motor adaptation, looking at cortical, sub-cortical and cerebellar contributions to actions.

This course is intended to provide the understanding and methodological tools required to design and build models of diverse experimental results in behavioural neuroscience, and also to embed modeling questions into the heart of the specification of new empirical questions that will decide between and refute models. This course also teaches critical thinking by linking theory with data, using models as tools.

Requirements

Calculus, reasonable knowledge of a high level computer programming language (python, R, julia, matlab); familiarity with probability theory, linear algebra, and information theory

Coursework

Each lecture will have 60 minutes of homework

There will be a final written exam.

Literature

Dayan P & Abbott LF (2001) Theoretical Neuroscience, MIT Press.

Sutton RS & Barto AG (2018) Reinforcement Learning: An Introduction, MIT Press. [2nd edition]

Zhaoping, L. (2014) “Understanding vision: theory, models, and data”, Oxford University Press (This book is available in the Unviersity of GTC library, there are learning support with short video lectures and quizzes at zhaoping.thinkific.com useful)

Lecture list:

1. Introduction (Zhaoping, Ilg, Dayan)

2. Classical conditioning: short- and long-term prediction in the brain (Dayan)

3. Instrumental conditioning: ongoing and learned choice of actions (Dayan)

4. Model-based and model-free methods of choice (Schwartenbeck)

5. Vigour, exploration, risk and more (Lloyd)

6. Efficient coding principle in vision (Zhaoping, learning support at zhaoping.thinkific.com)

7. Visual attention and visual saliency (Zhaoping, learning support at zhaoping.thinkific.com)

8. Visual decoding (Zhaoping, learning support at zhaoping.thinkific.com)

9. A new framework for understanding vision from the perspective of the primary visual cortex (Zhaoping, learning support at zhaoping.thinkific.com)

10. Feed-foward and feedback control (Ilg)

11. Motor adaptation (Ilg)

12. Reinforcement learning in motor control (Ilg)

13. Motor learning architectures (Ilg )