Models of Higher Brain Function

Course: Models of Higher Brain Function

Instructor: Prof. Dr. Sprekeler

Bernstein Center for Computational Neuroscience

This is a private repository for storing the assignments of the programming tutorial of the course Models of Higher Brain Function.

Brief description of each Assignment:

1. Deep Linear Networks:

In this assignment, we explore the learning dynamics of a deep linear network and contrast them with those of a shallow network.

2. Slow Feature Analysis:

We apply Linear Slow Feature Analysis (SFA) transformation on time-dependent signals and explore SFA on a high-dimensional correlated signal.

3. Visual Attention:

The goal of this assignment is to compute saliency maps for natural images, similar to the work presented in Itti, Koch & Niebur (1998).

4. Perceptual Bistability:

In this assignment, we delve into a model of binocular rivalry presented by Laing and Chow (2002) and a model of perceptual bistability developed by Moreno-Bote et al. (2007).

5. Decision Making I & II:

This assignment involves analyzing the reaction time distribution of the drift diffusion model for perceptual decision making. Additionally, we utilize the Chapman-Kolmogorov equation to find the time-dependent distribution of the decision variable in a drift-diffusion model for decision making. The Fokker-Planck equation is also explored.

6. Reinforcement Learning I & II:

In these assignments, we learn policies for the optimal strategy for a 10-armed bandit and implement and practice SARSA and Q-Learning algorithms in 2D grid games.

Final Project: Decision Making in Low Rank Recurrent Neural Network

For the final project, we will study the Low Rank RNN frameworks and replicate the results based on the paper by Debreuil et al.(2022). You can find the report of the project available through this link.