Thomas Pethick

Research Interests:

  • Bayesian Optimization
  • Machine Learning

Biography

I received my Master with Honours in Computer Science and Engineering from the Technical Danish University in the summer 2019. My master project titled “Gaussian Processes for Economic Models” was jointly supervised by Prof. Ole Winther and Prof. Volkan Cehver. I started my PhD at LIONS in 2019.

Publications with LIONS

Improving SAM Requires Rethinking its Optimization Formulation

W. Xie; F. Latorre; K. Antonakopoulos; T. M. Pethick; V. Cevher 

2024. 41st International Conference on Machine Learning (ICML 2024), Vienna, Austria, July 21-27, 2024.

Stable Nonconvex-Nonconcave Training via Linear Interpolation

T. M. Pethick; W. Xie; V. Cevher 

2023. Thirty-seventh Conference on Neural Information Processing Systems, New Orleans, Louisiana, USA, December 10-16, 2023.

Solving stochastic weak Minty variational inequalities without increasing batch size

T. M. Pethick; O. Fercoq; P. Latafat; P. Patrinos; V. Cevher 

11th International Conference on Learning Representations ICLR2023, Kigali, Rwanda, May 1-5, 2023.

Federated Learning under Covariate Shifts with Generalization Guarantees

A. Ramezani-Kebrya; F. Liu; T. M. Pethick; G. Chrysos; V. Cevher 

Transactions on Machine Learning Research. 2023. num. 06.

Finding Actual Descent Directions For Adversarial Training

F. Latorre; I. Krawczuk; L. T. Dadi; T. M. Pethick; V. Cevher 

2023. 11th International Conference on Learning Representations (ICLR), Kigali, Rwanda, May 1-5, 2023.

Revisiting adversarial training for the worst-performing class

T. M. Pethick; G. Chrysos; V. Cevher 

Transactions on Machine Learning Research. 2023. 

Escaping limit cycles: Global convergence for constrained nonconvex-nonconcave minimax problems

T. M. Pethick; P. Latafat; P. Patrinos; O. Fercoq; V. Cevher 

2022. 10th International Conference on Learning Representations (ICLR 2022), Virtual, April 25-29, 2022.

Subquadratic Overparameterization for Shallow Neural Networks

C. Song; A. Ramezani-Kebrya; T. Pethick; A. Eftekhari; V. Cevher 

2021. 35th Conference on Neural Information Processing Systems (NeurIPS 2021), Sydney, Australia, December 6-14, 2021.

Sifting through the Noise: Universal First-Order Methods for Stochastic Variational Inequalities

K. Antonakopoulos; T. M. Pethick; A. Kavis; P. Mertikopoulos; V. Cevher 

2021. NeurIPS 2021 : Thirty-fifth Conference on Neural Information Processing Systems, Sydney, Australia [Virtual only], December 6-14, 2021.

Access map