Projects
Some projects that I have created over the years.
Click on the title to view the project.
ScanBoard
A webtoy used to scan whiteboards fitted with ScanBoard fiducials.
Cricket Scores
A web-based scoreboard for cricket.
Tic-Tac-Toe Ai
This project implements a neural network that learns to play Tic Tac Toe using real backpropagation and epsilon-greedy exploration. The AI can play against a human or train by playing games against itself, gradually improving its strategy over thousands of iterations.
The network has a simple architecture: 9 input neurons (board positions), 18 hidden neurons, and 9 output neurons (move probabilities). Moves are chosen probabilistically, sometimes randomly to encourage exploration. After each game, the AI updates its weights and biases based on the outcome, allowing it to learn winning strategies over time.
All network parameters are saved to a file, enabling persistent training across sessions. Over multiple games, the AI demonstrates adaptive behavior, increasingly avoiding mistakes and making more strategic moves, illustrating how a neural network can learn even in a simple, deterministic game environment. Click HERE for a more detailed writeup.