Final Appraisal Seminar for Tenure Track Assistant Professor Sadegh Talebi

Tenure Track Assistant Professor Sadegh Talebi
Tenure Track Assistant Professor Sadegh Talebi

Join the Final Appraisal Seminar for Tenure Track Assistant Professor Sadegh Talebi. The seminar will feature presentations on research achievements during the tenure track period, future research directions, and reflections on teaching, followed by questions from external opponents and the appraisal committee.

Programme

15:00–15:20 Arrival
15:25–15:30 Short welcome and introduction by the Associate Dean for Research
15:30–16:15 45-minute presentation by the candidate:
  • Presentation of research conducted during the tenure track period and its relation to the state of the art (approx. 30 min)
  • Presentation of future research visions (approx. 7–8 min)
  • Reflections on teaching experience and future teaching development (approx. 7–8 min)
16:15–16:20 Short break
16:20–16:40 Research questions by external opponent, Professor Aurélien Garivier
16:40–17:00 Research questions by external opponent, Professor Gergely Neu
17:00–17:15 Questions on teaching experience and future teaching visions by Deputy Head for Education, Boris Düdder
17:15–17:30 Questions regarding future ideas, collaboration and international networking by Head of Department and Associate Dean for Research
17:30–17:45 Evaluation by Tenure Track Appraisal Committee (closed session)
17:45–18:00 Concluding evaluation and conversation between candidate and committee in front of audience
18:00–19:00 Reception

Title:

Provably Efficient Reinforcement Learning Beyond the Markov and Risk-Neutral Formulations

Abstract:

My research focuses on the theoretical foundations of reinforcement learning (RL), with the goal of developing learning algorithms with rigorous performance guarantees in terms of regret and sample complexity. My work advances the state of the art for classical RL while extending its theoretical foundations to settings where its standard assumptions do not hold. More recently, my research has focused on risk-sensitive RL, going beyond the standard risk-neutral objective by incorporating risk measures such as entropic risk. In this setting, I develop algorithms and theoretical guarantees for RL with risk-aware objectives. I have also investigated RL beyond the classical Markov assumption, extending these guarantees to environments where reward and transition dynamics are governed by an underlying (possibly latent) finite-state automaton, capturing memory and partial observability. My work spans both online and offline RL settings. I will present recent developments along these directions and conclude by presenting an overview of my teaching and service contributions to the department, together with my vision for future research on the theory of RL.