Self-supervised learning for inverse problems
Title
Self-supervised learning for inverse problems
Speaker
Julian Tachella - ENS Lyon
Abstract
In this talk, I will cover some concepts and recent advances in the emerging field of self-supervised learning methods for solving imaging inverse problems. Self-supervised learning is a fundamental tool deploying deep learning solutions in scientific and medical imaging applications where obtaining a large dataset of ground-truth images is either expensive or impossible. I will present recent self-supervised methods, which can be applied in settings where the noise distribution is partially unknown, and the forward operator is ill-posed/incomplete. Finally, I will discuss their theoretical underpinnings and present practical imaging applications.
Bio
Julián Tachella is a research scientist at CNRS (ENS Lyon), specializing in signal processing, machine learning, and computational imaging. He is a lead developer of DeepInverse, a widely used open-source library that has become a key tool for solving inverse problems with deep learning. His research advances include self-supervised methods and efficient neural networks for high-quality image reconstruction.
When
Friday June 12th, 10:30
Where
Room 322, UniGe DIBRIS/DIMA, Via Dodecaneso 35