Thomas Lee

About

Hi! I am a postdoc working on continual learning, decision making and time series forecasting. I work at Mila in the Chandar Research Lab, supervised by Sarath Chandar. I love to think about problems from a fundamental perspective, often starting from analysing a problem using Bayesian statistics and decision theory, then using these insights gained to propose practical algorithmic improvements. I am also very interested in boundedness and how it effects decision making.

Before my postdoc, I did a PhD supervised by Amos Storkey at Edinburgh University in the Bayesian and Neural Systems group. Prior to the PhD, I completed a master’s and bachelor’s in Computer Science at the University of Cambridge.

Publications

For a full list of publications check out my google scholar page.

Adapting Time Series Foundation Models through Data Mixtures [paper]
T. L. Lee, E. M. Ponti, and A. Storkey.
Pre-print, 2026.

Signature-Kernel Based Evaluation Metrics for Robust Probabilistic and Tail-Event Forecasting [paper]
B. R. Redhead, T. L. Lee, P. Gu, V. Elvira, and A. Storkey.
Pre-print, 2026.

Forgetting is Everywhere [paper]
B. Sanati, T. L. Lee, T. McInroe, A. Scannell, N. Malkin, D. Abel, and A. Storkey.
CoLLAs, 2026. (Oral)

Lightweight Online Adaption for Time Series Foundation Model Forecasts [paper]
T. L. Lee*, W. Toner*, R. Singh, A. Joosem, and M. Asenov.
ICML, 2025.

Performance of Zero-Shot Time Series Foundation Models on Cloud Data [paper]
W. Toner*, T. L. Lee*, A. Joosem, R. Singh, and M. Asenov.
ICLR, I Can’t Believe It’s Not Better Workshop, 2025.

Chunking: Continual Learning is not just about Distribution Shift [paper]
T. L. Lee, and A. Storkey.
CoLLAs, 2024. (Oral)

Approximate Bayesian Class-Conditional Models under Continuous Representation Shift [paper]
T. L. Lee, and A. Storkey.
AISTATS, 2024.

Hyperparameter Selection in Continual Learning [paper]
T. L. Lee, S. P. Hellan, L. Ericsson, E. J. Crowley, and A. Storkey.
CoLLAs Workshop, 2024.

On Overcompression in Continual Semantic Segmentation [paper]
M. Kowalski, T. L. Lee, and A. Storkey.
Pre-print, 2022.