Second Workshop on the Convergence of Large Scale Simulation/HPC and Artificial Intelligence
Second Workshop on the Convergence of Large Scale Simulation/HPC and Artificial Intelligence
June 20, Frankfurt Germany
in conjunction with ISC High Performance 2019
In 2016, an artificial intelligence (AI) named AlphaGo became the first machine to beat a human 9-dan professional Go player on a full-sized board. Since then, AI has proven its potential to revolutionize our social life, how we live, work, learn, discover, and communicate. But does AI also have the potential to fundamentally change how we do science? The recent stream of publications suggests it may, but there is still much work to be done before AI can enter the day-to-day toolbox of scientists.
This workshop will discuss ways that Deep Learning and Artificial Intelligence can be combined with traditional high performance computing applications to accelerate the pace of scientific discovery from High Energy Physics to Life Sciences and Healthcare. One traditional paradigm in scientific computing uses large scale simulation at the core, where data analytics is used mainly for pre and post processing of the data. In recent years, HPC applications have started to use data analytics and AI also on a more cooperative basis where the strengths of each converge to form a powerful new tool for science. In this workshop we will discuss both uses of AI, traditional and new.
Agenda
Agenda 2nd Workshop on Convergence of Large Scale Simulation/HPC and Artificial Intelligence
Time : Thursday, June 20th 9am - 1pm
Location : Kilobyte
09:00 – 09:15 am Welcome and Introduction (Christoph Angerer / Axel Koehler , NVIDIA)
09:15 – 09:50 am Keynote 1: Learning to Optimize, Search and Plan (Azalia Mirhoseini, Google Brain) (Bio) (Abstract)
09:50 – 10:15 am Modeling Turbulent/reactive Flows using High Performance Computing Data and Deep Learning (Mathis Bode, RWTH Aachen)(Slides)
10:15 – 10:40 am Enhancing GPU-accelerated Plasma Simulations via Deep Neural Networks (Nico Hoffmann, Helmholtz-Zentrum Dresden-Rossendorf)(Slides)
10:40 – 11:15 am Break
11:15 – 12:00 pm Keynote 2: Learnable Representations of Code (Tal Ben-Nun, ETH Zurich) (Bio) (Abstract)(Slides)
12:00 – 12:25 pm Learning to Learn and Learning to Optimize for High-Throughput Hyperparameter Search using HPC (Alper Yegenoglu, Jülich Supercomputing Centre)(Slides)
12:25 – 12:50 pm Accelerating Radiative Transfer Calculations using Neural Networks (Menno A. Veerman, Wageningen University and Research, SURFsara)(Slides)
12:50 - 01:00 pm Wrapping Up
01:00 pm End
Call for Contributions
We invite submissions of high-quality, novel, and original research results on applying artificial intelligence technologies to HPC applications, especially large scale simulations, and the following related topical areas:
- HPC*AI Tools and Methodologies for the design, implementation, and verification of converged applications
- Scaling AI for large scale simulations (training and inference)
- Employing AI (ML/DL) for scientific computing applications (esp. large scale simulations, but also other areas are welcome)
- Tools and methodologies for integrating AI into HPC applications
- Use cases and success stories for converged applications
Important Deadlines
- Extended abstract submission deadline: Wednesday, April 24st, 2019
- Acceptance Notification: Friday, April 26th, 2019
- Workshop date: Thursday, June 20th, 2019
Submission Instructions
Extended abstracts should be mailed to the workshop organizers (cangerer@nvidia and akoehler@nvidia.com) by April 24, 2019. Extended abstracts should be 1-4 pages long and contain references, comparisons to related work, and results if applicable. Abstracts should be formatted according to Springer's single column LNCS style (see http://www.springer.com/computer/lncs?SGWID=0-164-6-793341-0 for LaTeX and Word templates).
Program Committee
Christoph Angerer, NVIDIA (organizer)
Axel Köhler, NVIDIA (organizer)
Torsten Hoefler, ETH Zurich
Jean-Roch Vlimant, California Institute of Technology
Peter Dueben, ECMWF
Yu Wang, LRZ
Mathis Bode, RWTH Aachen