Set of gamified scenarios where we implement various AI algorithms.
We implement tree search using iterative deepening with alpha-beta pruning in order to estimate the best move for a particular state. We also implement state hashsing and use it for repeated state checking which greatly improves the overall performance.
We explore Hidden Markov Models and algorithms to infere probability distributions over the hidden states and observations. We also implement ways of estimating the most likely sequences of observations for a given model and use it to perform classification.
We implement the Q-learning algorithm and train it using linear scheduling epsilon greedy policy. We then investigate the hyperparameters of the model and their effect on its performance.
