Context
This project was completed during my time as a Computational Neuroscience Researcher at the Ma Lab for Computational Neuroscience.
Scope & Methodology
In short, it’s an implementation of a restricted boltzmann machine geared towards simulating a visual occlusion decision-making task. In this case, the decision is whether a line, blocked by some occlusion, is perceived as a singular line, or two separate lines.
The Restricted Boltzmann Machine is created in order to analyze different types of probabilistic models, and determine which is closest to the results seen in human trials. More information on the Restricted Boltzmann Machine and contrastive divergence learning can be found in this great lecture. More information on the task being replicated can be found at this link.
The project was originally implemented in Matlab, and used as materials for an internal presentation, which you can find on my GitHub (check the links below).