I am currently a Postdoctoral Scholar at the Research Center for Information Technology Innovation (CITI) at Academia Sinica (AS) in Taiwan.

I received my Ph.D. degree in Computer Science in 2023, where I was part of the Machine Learning Section of DIKU (Department of Computer Science, University of Copenhagen), under the supervision of Yevgeny Seldin. I am also a member of the DeLTA Lab (Denmark Learning Theory and Applications). Right after, I joined as a Postdoc Researcher in the Department of Mathematics and Computer Science (IMADA) at South Denmark University (SDU), where I was a member of the Data Science and Statistics (DSS) Group under IMADA. In the meantime, I was also affiliated with the Learning Theory and Optimization Collaboratory of Pioneer Center for Artificial Intelligence (PI) in Denmark.

I received my BSc degree in Physics from National Taiwan University (NTU) in 2017. After my BSc, I worked as a research assistant at the Institute of Information Science at Academia Sinica, under the supervision of Chi-Jen Lu in online learning and lifelong learning, and wrote a few blog posts on different topics (no longer maintained).

My research targets fundamental problems in Machine Learning, aimed at identifying and resolving the structural bottlenecks of learning algorithms. My current focus is Reinforcement Learning (RL), complementing my background in statistical learning theory (specifically for randomized predictors) and sequential decision-making (bandits and lifelong learning).

 Recent Work

• ObjectRL: An Object-Oriented Reinforcement Learning Codebase [paper]
  Gulcin Baykal, Abdullah Akgül, Manuel Haussmann, Bahareh Tasdighi, Nicklas Werge, Yi-Shan Wu, Melih Kandemir
  preprint.

• Improving Actor-Critic Training with Steerable Action-Value Approximation Errors [paper]
  Bahareh Tasdighi, Nicklas Werge, Yi-Shan Wu, Melih Kandemir
  ECAI 2025.

• Deep Exploration with PAC-Bayes [paper]
  Bahareh Tasdighi, Manuel Haussmann, Nicklas Werge, Yi-Shan Wu, Melih Kandemir
  ECAI 2025

• Recursive PAC-Bayes: A Frequentist Approach to Sequential Prior Updates with No Information Loss [paper]
  Yi-Shan Wu, Yijie Zhang, Badr-Eddine Chérief-Abdellatif, Yevgeny Seldin
  NeurIPS 2024 (spotlight)

 News

• Jan 2026: I have been awarded a fellowship under the Academia Sinica Postdoctoral Scholar Program and will be joining CITI, Academia Sinica.

• August 2025: I will be attending RLC in Alberta.

• July 2025: I will be attending DLRL summer school organized by AMII in Alberta.

• December 2024: I will be attending NeurIPS 2024 in Vancouver.

• October 2024: I will attend the D3A 2.0 event in Nyborg, where I will co-organize the "Machine Learning Theory" session, as well as give a talk at the "Bayesian Methods for Uncertainty Quantification" session.

• May 2024: I will be attending AISTATS 2024 in Valencia.

• Feburary2024: I will attend the D3A 1.0 event in Nyborg, where I will co-organize the "Machine Learning Theory" session.

• October 2023: I will be co-organizing AI Research Moonshots Workshop at Pioneer Centre for Artificial Intelligence in Copenhagen.

• September 2023: I will be attending GenU 2023 in Copenhagen.

• September 2023: I will be participating EWRL 2023 in Brussels.

• April 2023: I will be attending AISTATS 2023 in Valencia.

• March 2023: I will be visiting CWI, Amsterdam, for a month.

• December 2022: I will be attending NeurIPS 2022 in New Orleans.

• November 2022: I will be visiting CWI, Amsterdam, for a month.

• July 2022: I will be attending COLT 2022 in London.

• April 2022: I will be visiting the statistics department of UCL, London, for two weeks.

• November 2021: I will be attending 1st NordicAIMeet in Oslo.

• October 2021: I will be visiting CWI, Amsterdam for a week.