Teaching at TU Delft

• Q1: Reinforcement learning lecture in the Fundamentals of AI Programme (IFEEMCS520100)
  Introduction into basic reinforcement learning. The goal is to introduce non-computer-science students to the basic concepts, hands-on experience with tabular Q-learning, how modern algorithms extend this with neural networks and a general overview over the strengths and weaknesses of RL. You can watch the recorded mini-lectures of this course below.
• Q3: Bachelor course Automata, Computability and Complexity (CSE2315)
  This course introduces three areas of the theory of computation.
  • First, automata are used to recognize words in (formal) languages, and we will discern several types of languages along with the types of automata that can recognize them.
  • Second, once we have an intuition about languages, we discuss the topic of computability, where you will learn not only what kinds of problems a computer can solve but also how to prove this.
  • Third, we will examine the class of computable problems and make a distinction between "easy" and "hard" problems and delve into the famous problem: P = NP?
• Q4: Master course Deep Reinforcement Learning (DSAIT4115)
  A deep dive into theory and application of modern model-free deep reinforcement learning (RL) methods. Course contents are:
  • deep learning methodology and architectures
  • stabilization of approximated value estimation
  • modern actor-critic methods
  • planning as inference and robust RL
  • exploration with deep networks
  • offline reinforcement learning
  • deep multi-agent reinforcement learning
  • applied multi-task RL
• Q4: Research projects for the Bachelor (CSE3000) and Master students (CS4210-B)
  Hands-on projects about current research topics for groups of 3-5 students.
  • Past projects for Bachelor students (CSE3000):
    • Generalization of Visual Transformers (Assini, 2022; Colle, 2022; de Kruijff, 2022).
    • The attention LSTM model (Acquati, 2021; van Diepen, 2021; Sina, 2021 Wang, 2021).
  • Past projects for Master students (CS4210-B):
    • Lights out: training RL agents robust to temporary blindness (Ordonez, Tromp & Julbe, 2023).
    • New Techniques for Fast Convergence in Reinforcement Learning.
    • High-confidence Monte Carlo Tree Search.
    • A Milestone-based Adaptive Exploration Rate Approach for Reinforcement Learning.
    • Hidden State Based Uncertainty Estimation using Ensemble LSTMs.
    • Pessimistic Exploitation for Optimistic Exploration in Reinforcement Learning.

A short introduction into Reinforcement Learning

If you are interested in RL, but do not know what it exactly is, you can watch the following videos, which are lectures from the course "Fundamentals of AI Programme" (IFEEMCS520100).