Advanced computational imaging: the rise and fall of deep neural nets

Sustainable Scientific Computing

High-performance computing is an essential tool for modern scientific development thanks to its data and compute intensive nature, allowing scientists to perform increasingly complex analyses in the ongoing information tsunami. While challenges in data acquisition and processing can often hinder HPC from reaching its full potential, scientific consortia's desire for energy-efficient hardware/software co-design and green computing push the limits of HPC. In this talk, we will present some of these challenges in two specific examples from astrophysics and nuclear fusion.

Gilles Fourestey is the director of operation of the HPC platform and the Eurofusion Advanced Computing Hub at EPFL. He has 20 years of experience in High Performance software development, code optimization and modernization for CPUs and GPUs, and data analysis in several distinct science fields, such as astrophysics, material science, fusion and precision medicine. He also has extensive expertise in numerical analysis, mathematical optimization and optimal control. His background is applied mathematics and scientific computing.