Lecture Program
04/03/2025
1. Basics of RL (Patrizi):MDP, value and policy iteration
Slides | Video of Lecture 1 and 2
2. Markov RL (Iocchi): exploration/exploitation, Q-learning, Sarsa
05/03/2025
3. Linear Temporal Logic (De Giacomo): LTLf, DFA
4. Non Markov Reward Decision Process (Patrizi): NMRDP, restraining bolts, reward machines
07/03/2025
5. Markov Deep RL (Capobianco): deep learning, DDQN, Actor Critic methods
11/03/2025
6. Automata learning (Paludo Licks): L*, merging states methods, learning automata for non-Markov RL
7. Deep Learning on Sequential Data and Non Markov Deep RL (Umili): Deep learning for sequential data (models, applications), hybrids between RNN and DFA (L* extraction, DeepDFA), non- markov deep RL
12/03/2025
8. Neurosymbolic (NeSy) AI (Umili, Marconato): introduction and applications, grounding symbols, Logic Tensor Network (LTN), LTN for LTLf, Reasoning Shortcuts
9. NeSy RL (Umili, Hyde): noisy symbols, Neural Reward Machines, unremovable reasoning shortcuts for DFAs, towards learning perceptions and rules from data, hidden triggers
14/03/2025
10. Other challenges (Umili): Generalizing to unseen formulas, towards using natural language in place of LTL, RL and LTL for autoregressive sequence generation