Code repository for the self-initiated group research project "An Investigation of Reinforcement Learning Algorithms for Mastering the Game of Santorini", completed as part of an Independent Study Module during Semester 2 of AY20/21 at the National University of Singapore.
In this project, we investigated various reinforcement learning based strategies for an agent to effectively learn and play the board game Santorini. These strategies include:
- Minimax search + linear value function approximators trained using reinforcement learning techniques (RootStrap and TreeStrap)
- Non-linear value function approximation using Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) + Monte Carlo Tree Search