AI Research: Bayesian ML, Causal ML, RL, Graph ML

About the course

This course introduces four research directions that are advancing rapidly within modern AI: Bayesian ML, Causal ML, Reinforcement Learning, and Graph ML. Each direction is examined through its theoretical foundations, open research questions, and connections to real-world problems. The course lays the groundwork for choosing a Master's thesis topic.

What you'll learn

✓Derive the ELBO and implement a variational autoencoder (VAE) for a generative task

✓Construct a causal graph and apply do-calculus to estimate interventional effects

✓Train an RL agent with PPO on a benchmark task and analyze its convergence

✓Implement a GNN with message passing and assess its expressive power

Key topics

•Bayesian ML: variational inference (VI), MCMC, deep Bayesian models (BNN, VAE)

•Causal ML: causal graphs (DAGs), do-calculus, counterfactual reasoning

•Causal inference from observational data: propensity scores, instrumental variables

•Reinforcement Learning: MDPs, policy gradient (REINFORCE, PPO), Q-learning (DQN)

•RL for LLMs: RLHF, GRPO, reward hacking

•Graph Neural Networks: GCN, GAT, message passing, expressive power (WL test)

•Geometric deep learning: invariance and equivariance

•Cross-cutting connections: Bayesian RL, causal representation learning

This description was generated automatically and has not yet been reviewed by an instructor — it's a draft for discussion.

About the course

What you'll learn

Key topics

Feedback