Sam Kessler

• I’m joining Microsoft Cambridge as a Post-doc.
• I did my PhD student in the Machine Learning Research Group under the supervision of Steve Roberts. My thesis is entitled “Retaining skills under distribution shifts: Bayesian continual learning, continual reinforcement learning, and applications” (to be made public soon).

News

• I am joining Microsoft Cambridge as a Post-doc thinking about LLM efficiency; how to make the LLM inference and training more efficient.
• I am reviewing for Neurips 24.

Publications

Preprints

• Kessler, Samuel*, Le, Tam*, Nguyen, Vu. “SAVA: Scalable Learning-Agnostic Data Valuation”, under submission [arXiv], [code].
• Davis, Oscar, Kessler, Samuel, et al. “Fisher Flow Matching for Generative Modeling over Discrete Data”, under submission [arXiv].

Published papers

• Kessler, Samuel, et al. “The Effectiveness of World Models for Continual Reinforcement Learning”, in CoLLAs 2023 [arXiv], [code].
  Also appeared at the Deep Reinforcement Learning workshop @ Neurips 2022.

• Woods, Kieran* and Kessler, Samuel*, et al. “Few-shot Learning Patterns in Financial Time-Series for Trend-Following Strategies”, in Journal of Financial Data Science [arXiv].

• Kessler, Samuel, et al. “On Sequential Bayesian Inference for Continual Learning?”, in Entropy 2023 [Entropy], [code].
  Also appeared at the Advances in Approximate Bayesian Inference workshop 2022.

• Kessler, Samuel et al. “Same State, Different Task: Continual Reinforcement Learning without Interference”, oral presentation in AAAI 2022, [arXiv], [code].
  Continual learning workshop @ ICML, 2020.

• S. Kessler, B. Thomas, S. Karout. “Continual-wav2vec2: an Application of Continual Learning for Self-Supervised Automatic Speech Recognition”, in ICASSP 2022, [arXiv].
  Self-supervised learning workshop for reasoning and perception @ ICML 2021.

• B. Thomas, S. Kessler, S. Karout. “Efficient Adapter Transfer of Self-supervised Speech Models for Automatic Speech Recognition”, in ICASSP 2022, [arXiv].

• J. Hergueux*, S. Kessler*. “Follow the Leader: Technical and Inspirational Leadership in Open Source Software”, in ACM SIGCHI 2022 [arxiv].

• Kessler, Samuel, et al. “Hierarchical Indian Buffet Neural Networks for Bayesian Continual Learning”, spotlight in UAI 2021, [arXiv] [code].
  Neurips 2019 Bayesian Deep Learning Workshop.

Workshop papers

• Kessler, Samuel* and Salas, Arnold* et al. “Practical Bayesian Learning of Neural Networks via Adaptive Optimisation Methods”, in ICML 2020 Workshop on Uncertainty and Robustness in Deep Learning, [arXiv], [code].

*equal contribution

Teaching

• Teaching Assistant HT 2021, Applied Machine Learning, Graduate Course, Oxford Univeristy, Oxford Internet Institute.
• Teaching Assistant MT 2020, Machine Learning, Graduate Course, Oxford Univeristy, Oxford Internet Institute.

Find me

skessler{at}robots{dot}ox{dot}ac{dot}uk/Google Scholar/GitHub/Twitter/Linkedin/CV.