> > Workshops > Workshop 8
Workshop 08 - 2:00 to 5:30 pm
High performance AI in the industry
Chaired by Cristel Saudemont, France Director, Supercomputing & AI , Higher Education and Research and Frédéric Parienté, senior manager in the solutions architecture and engineering group, Nvidia
What can AI do for High-Performance Computation?
By Corentin Lapeyre, Research Scientist, Coordinator of the Helios workgroup, CERFACS
While machine learning, and more specifically deep learning, have yielded important breakthroughs in the past decade, (e.g. in image recognition, playing board games, or natural language processing), similarly notable contributions to modelling of physical processes are still rare. Notably, the field of high-fidelity simulation stands to benefit largely from accelerated time to solution using these techniques, but also tolerates little error in its predictions, making their application all the more challenging. As the field matures though, many hybrid simulation strategies are emerging that could help accelerate HPC simulation, all while managing the impact on accuracy. In this talk, several hybrid strategies that leverage both the advantages of classical high-performance fluid simulations and of deep learning will be explored, demonstrating some promising avenues for their adoption in the field of HPC.
|
Biography : Corentin Lapeyre is a research scientist at Cerfacs. After a PhD and postdoc on CFD of unsteady turbulent combustion phenomena, he joined Cerfacs permanently and started a workgroup on hybrid CFD and Artificial Intelligence-driven approaches to solving fluid flow problems, on topics including aerodynamic and reactive turbulent flows for aerospace applications, combustion-related risk and safety assessments, climate modelling, and oil reservoir modelling. |
|