Cyrille W. Combettes

Bio

I am a quantitative researcher at Capital Fund Management in Paris. Prior to this, I was a postdoc working with Jérôme Bolte. I obtained a Ph.D. in Machine Learning from Georgia Tech in May 2021, where I was advised by Sebastian Pokutta and worked on developing efficient algorithms for constrained optimization.

Publications

The iterates of the Frank-Wolfe algorithm may not converge
J. Bolte, C. W. Combettes, and É. Pauwels
Mathematics of Operations Research, to appear
[paper]

Revisiting the approximate Carathéodory problem via the Frank-Wolfe algorithm
C. W. Combettes and S. Pokutta
Mathematical Programming, series A, 197(1):191–214, 2023
[paper] [code] [slides]

Complexity of linear minimization and projection on some sets
C. W. Combettes and S. Pokutta
Operations Research Letters, 49(4):565–571, 2021
[paper] [code]

Projection-free adaptive gradients for large-scale optimization
C. W. Combettes, C. Spiegel, and S. Pokutta
Preprint arXiv:2009.14114, 2020
[paper] [code] [blog]

Boosting Frank-Wolfe by chasing gradients
C. W. Combettes and S. Pokutta
International Conference on Machine Learning, 119:2111–2121, 2020
[paper] [code] [blog] [slides] [video]

Blended matching pursuit
C. W. Combettes and S. Pokutta
Neural Information Processing Systems, 32:2044–2054, 2019
[paper] [code] [slides] [poster]

Talks

Boosting Frank-Wolfe by chasing gradients
INFORMS Annual Meeting
Online, October 2021

Boosting Frank-Wolfe by chasing gradients
International Conference on Operations Research
Online, September 2021

The Frank-Wolfe algorithm: Projection-free and sparsity
MAI Division Seminar, Zuse Institute Berlin
Online, April 2021
[slides]

Boosting Frank-Wolfe by chasing gradients
International Conference on Machine Learning
Online, July 2020
[web] [slides] [video]

Blended matching pursuit
Conference on Neural Information Processing Systems
Vancouver (Canada), December 2019
[web] [poster]

Revisiting the approximate Carathéodory problem via the Frank-Wolfe algorithm
Conference on Data Science and Optimization
Toronto (Canada), November 2019
[web] [slides]

Blended matching pursuit
Conference on Discrete Optimization and Machine Learning
Tokyo (Japan), July 2019
[web] [slides]

Education

Georgia Institute of Technology, 2018–2021
Ph.D. in Machine Learning
- Thesis: Frank-Wolfe Methods for Optimization and Machine Learning
  [pdf] [slides]
- Committee: Alexandre d'Aspremont, Swati Gupta, Guanghui Lan, Arkadi Nemirovski, Sebastian Pokutta (advisor)

Massachusetts Institute of Technology, 2017–2018
Master of Business Analytics

École Centrale Paris, 2015–2018
M.Sc. in Applied Mathematics

Lycée Louis-le-Grand, 2009–2015

Misc

Soccer Intramurals
- 2-time winner (Centrale 2016, 2017)
- 2-time finalist (MIT 2018, Georgia Tech 2020)

Dean's Fellowship, MIT, 2017
In recognition of outstanding academic record, personal achievements, and professional promise

Jean Gaillard Memorial Fellowship, École Centrale Paris, 2017