This repo contains my projects in Probability methods in AI course taken in HUJI.
EX1: Hidden Markov Model
This exercise focuses on building an HMM class and predicting the Prior, Likelihood, and Posterior.
After that, we use the model to identify corrupted data.
EX2: Sampling-Based Inference
In this project, I reused the HMM class from EX1 to perform sampling-based inference, implementing both Gibbs Sampling and Likelihood Weighting.
EX3: Parameter Learning
In this exercise, we move from using predefined CPDs to learning model parameters from data.
Two learning methods were applied:
🔹MLE (Maximum Likelihood Estimation) over complete data.
🔹EM (Expectation-Maximization) for learning with missing data.
EX4: Reinforcement Learning
This project applies reinforcement learning techniques to solve a maze game called “The Fish Pond”.
We implemented the Q-learning algorithm in both offline and online learning settings.