Module 1. Modeling Behaviour (5 lectures)

• The Rescorla-Wagner Rule + Exploration-Exploitation (Boris)
• Reinforcement Learning: TDRL & Model-Based Approaches (Boris)
• Reinforcement Learning as Rational Decision-Making (Joao)
• Perceptual Decision-Making: Drift Diffusion Model (Boris)
• Inferring Latent Behavioral States (Heike Stein)

Module 2. Encoding/Decoding Models of Neural Data (3 lectures)

• Introduction to Neural Data: Encoding/Decoding Dichotomy (Joao)
• Neural Encoding (Joao)
• Neural Decoding (Joao)

Module 3. Network Models of Cognition (4 lectures)

• Building Blocks: Single Neuron Models (Pierre Houzelstein)
• Feed-Forward vs. Recurrent Networks (Joao)
• Linear Dynamical Systems (Joao)
• Modeling Complex Cognitive Tasks (Joao)

Drop me an email if you want the slides for my lectures.

Course Content (2 lectures)

• Biophysical models: from synapses to EEG
• Inferring network models from data

Drop me an email if you want the slides for my lectures.

Lecture: Use of recurrent neural networks to study cognition

After a brief reminder on RNNs, this course introduced and explained the notions of state space, dimensionality reduction, and dynamic system. It presented a framework to reverse engineer a recurrent neural network to gain insights on their working. The course provided a comparison with neuronal data and various examples of applications: decision-making, working memory, spatial navigation (head direction and grid cells), motor cortex.

Drop me an email if you want the slides for my lectures.

Tutorial Overview

A hands-on tutorial on understanding and implementing low-rank recurrent neural networks for modeling cognitive tasks. The tutorial covers:

• Mathematical foundations of low-rank networks
• Applications to working memory and decision-making tasks
• Analysis techniques for understanding network dynamics
• Reverse engineering trained RNN

Resources:

• GitHub Repository with code and notebooks