An offline-friendly Electron app that lets scholars highlight, tag, and instantly resurface PDF passages across multiple research projects.
An investigation into the behavior of gradients durning training of a simple CNN image classifier and how it relates to generalization.
From-scratch implementations of core machine learning and statistical algorithms including K-Means, Random Forest, Naive Bayes, Kalman Filter, and MCMC methods.
Neural networks trained to classify NFL formations and plays from Madden gameplay footage using computer vision techniques.
Hacking a lunar lander game with an automated PID controller to get perfect landings each time.
A neural network trained to play Atari Breakout. Instead of using the traditional Q-learning technique, we implemented Yann LeCun's world model architecture combined with Monte Carlo Tree Search