Function approximation problems in signal processing and deep learning

/medias/photo/seminaire-dit_1626769502506-jpg

Computing with just the right amount of resources has always been a key problem in both hardware and software-oriented settings. In this context, energy efficient computing is an issue that is becoming more and more critical as the slowdown and eventual end of Moore’s law is in sight.

The goal of my research is to aid the development of energy-efficient algorithms and systems by attacking the problem at the numeric/algorithmic level. This leads to approaches that are usually application-specific, but which rely on tools from various domains such as approximation theory, computer arithmetic, numerical analysis, number theory and convex and integer optimisation.

In this talk I will present how these different domains come together to solve resource-efficient design problems in signal processing and machine learning.