Hello! I’m a postdoc at Harvard University, hosted by Heng Yang. Previously I did my phd at Boston University, under the amazing guidance of Yannis Paschalidis and Ashok Cutkosky. Long ago I did my undergrad at Tsinghua University.

I’m broadly interested in the foundation of data science. Specifically, my research centers around adaptive online learning, which concerns the theory and practice of sequential decision making with Bayesian-type prior knowledge.

On the application side, I’m excited about various aspects of robotics and automation, especially algorithmic approaches that efficiently utilize large-scale pretraining (e.g., continual fine-tuning and conformal prediction).

Email address: zhiyuz [at] seas (dot) harvard (dot) edu

CV   Github   Google Scholar

Publication

Five representative works

• The Benefit of Being Bayesian in Online Conformal Prediction
  ZZ, Zhou Lu, Heng Yang.
  Preprint.

• Fast TRAC: A Parameter-Free Optimizer for Lifelong Reinforcement Learning
  Aneesh Muppidi, ZZ, Heng Yang.
  NeurIPS 2024.

• Improving Adaptive Online Learning Using Refined Discretization
  ZZ, Heng Yang, Ashok Cutkosky, Ioannis Paschalidis.
  ALT 2024.

• Unconstrained Dynamic Regret via Sparse Coding
  ZZ, Ashok Cutkosky, Ioannis Paschalidis.
  NeurIPS 2023.

• Optimal Comparator Adaptive Online Learning with Switching Cost
  ZZ, Ashok Cutkosky, Ioannis Paschalidis.
  NeurIPS 2022. Also presented at ICML 2022 workshop.

Other works

• Sparsity-Based Interpolation of External, Internal and Swap Regret
  Zhou Lu, Y. Jennifer Sun, ZZ. (alphabetical order)
  Preprint.

• Population Dynamics Control with Partial Observations
  Zhou Lu, Y. Jennifer Sun, ZZ. (alphabetical order)
  Preprint.

• Adapting Conformal Prediction to Distribution Shifts Without Labels
  Kevin Kasa, ZZ, Heng Yang, Graham Taylor.
  Preprint.

• Discounted Adaptive Online Learning: Towards Better Regularization
  ZZ, David Bombara, Heng Yang.
  ICML 2024.

• Understanding Adam Optimizer via Online Learning of Updates: Adam is FTRL in Disguise
  Kwangjun Ahn, ZZ, Yunbum Kook, Yan Dai.
  ICML 2024.

• PDE-Based Optimal Strategy for Unconstrained Online Learning
  ZZ, Ashok Cutkosky, Ioannis Paschalidis.
  ICML 2022.

• Adversarial Tracking Control via Strongly Adaptive Online Learning with Memory
  ZZ, Ashok Cutkosky, Ioannis Paschalidis.
  AISTATS 2022.

• Provable Hierarchical Imitation Learning via EM
  ZZ, Ioannis Paschalidis.
  AISTATS 2021. Also presented at ICML 2020 Workshop.

PhD Dissertation.  Temporal Aspects of Adaptive Online Learning: Continuity and Representation