A GPU accelerated simulation of grid cells. Heavily inspired by Burak and Fiete et al. (2009) Paper.
This implementation demonstrates the potential for grid cells to detect velocity of input stimuli via feedback (reading out the velocity is not implemented).
Created as a final project for Projects in the Science of Intelligence (9.58) at MIT. In paper.pdf is the admitedly rough and oddly structured paper detailing this work further (Wells Crosby's version).
Run grid_cells.ipynb to see an interactive grid cell simulation. An Nvidia GPU will allows the program to run extremely quickly!
- Improving kernel generation
- Improving performance
- Use activation sum to create velocity detector
- Break global velocity into seperate cardianal directions
I will likely not work on this project further so please give those improvements a try!