About me

I am a researcher in the parietal team @ INRIA Saclay, since autumn 2019, working on unsupervised learning for time series and on deep learning methods applied to solving inverse problems.

My research interests touch several areas of Machine Learning, Signal Processing and High-Dimensional Statistics. In particular, I have been working on Convolutional Dictionary Learning, studying both their computational aspects and their possible application to pattern analysis. I am also interested in theoretical properties of learned optimization algorithms such as LISTA, in particular for the resolution of inverse problems.

I did my PhD under the supervision of Nicolas Vayatis and Laurent Oudre, in the CMLA @ ENS Paris-Saclay. My PhD focuses on convolutional representations and their applications to physiological signals. I continued working on unsupervised learning for time series with application to electrophysiological recordings during a 1,5 year in the Parietal Team. I am also involved in open-source projects such as joblib or loky, for parallel scientific computing, and benchopt, for reproducible benchmarks in optimization.

Latest publication and projects

Learning to solve TV regularized problems with unrolled algorithms
Cherkaoui Hamza; Sulam Jeremias; Moreau Thomas, Dec 2020, In proceedings of Advances in Neural Information Processing System
In this paper, we accelerate such iterative algorithms by unfolding proximal gradient descent solvers in order to learn their parameters for 1D TV regularized problems. While this could be done using the synthesis formulation, we demonstrate that this leads to slower performances. The main difficulty in applying such methods in the analysis formulation lies in proposing a way to compute the derivatives through the proximal operator.
Total Variation (TV) is a popular regularization strategy that promotes piece-wise constant signals by constraining the`1-norm of the first order derivative of the estimated signal. The resulting optimization problem is usually solved using iterative algorithms such as proximal gradient descent, primal-dual algorithms or ADMM. However, such methods can require a very large number of iterations to converge to a suitable solution. In this paper, we accelerate such iterative algorithms by unfolding proximal gradient descent solvers in order to learn their parameters for 1D TV regularized problems. While this could be done using the synthesis formulation, we demonstrate that this leads to slower performances. The main difficulty in applying such methods in the analysis formulation lies in proposing a way to compute the derivatives through the proximal operator. As our main contribution, we develop and characterize two approaches to do so, describe their benefits and limitations, and discuss the regime where they can actually improve over iterative procedures. We validate those findings with experiments on synthetic and real data.
Learning to optimize with unrolled algorithms slides
15 Apr 2021, At ML-MTP seminar - Montpellier
In this talk, I will first review how one can design unrolled algorithms to solve the linear regression with l1 or TV regularization, with a particular focus on the choice of parametrization and loss. Then, I will discuss the reasons why such procedure can lead to better results compared to classical optimization, with a particular focus on the choice of step sizes.
When solving multiple optimization problems sharing the same underlying structure, using iterative algorithms designed for worst case scenario can be considered as inefficient. When one aim at having good solution in average, it is possible to improve the performances by learning the weights of a neural networked designed to mimic an unfolded optimization algorithm. However, the reason why learning the weights of such a network would accelerate the problem resolution is not always clear.
In this talk, I will first review how one can design unrolled algorithms to solve the linear regression with l1 or TV regularization, with a particular focus on the choice of parametrization and loss. Then, I will discuss the reasons why such procedure can lead to better results compared to classical optimization, with a particular focus on the choice of step sizes.

Linked papers:
  1. Moreau & Bruna. Understanding Neural Sparse Coding with Matrix Factorization. ICLR 2017.
  2. Ablin, Moreau, Massias & Gramfort. Learning step sizes for unfolded sparse coding. NeurIPS 2019.
  3. Cherkaoui, Sulam & Moreau. Learning to solve TV regularised problems with unrolled algorithms. NeurIPS 2020.
BenchOpt 1.1 Apr 2021
Benchmarking tool for optimization.
BenchOpt is a package to simplify, make more transparent and more reproducible the comparisons of optimization algorithms.

BenchOpt is written in Python but it is available with many programming languages. So far it has been tested with Python, R, Julia and compiled binaries written in C/C++ available via a terminal command. If it can be installed via conda it should just work!