As part of the technical program of keynote speakers and presentation of peer-reviewed papers in seven technical sessions, HiPC 2017 has organized two plenary panels to be held on Days 2 and 3 of the conference. 

Each panel with offer attendees the chance to hear from experts in a popular area of high performance computing, data, and analytics and also to learn of developments in the field.

Below are brief descriptions of each panel. We will be posting additional information, including panel members, so please return to this page in the coming weeks.

Plenary Panel 1: 

HPC in India 

When : Tuesday, 19th December (5:30 PM -7:30 PM)

Description: 

As a special feature of HiPC 2017, a session on HPC initiatives in India is being organized and will feature presentations by leading researchers from academia, industry and Indian government R&D labs. 

Panelists: 

Plenary Panel 2: 

Machine Learning and Mathematical Modeling – Towards Robust, Reliable and Certifiable Machine Learning 

When : Wednesday, 20th December (3:30 PM – 5:30 PM)

 Description: 

For centuries, much of science and engineering has focused on transforming scientific insight into mathematical models (“the equations”) that have enabled effective tools for prediction and interpretation. These have enabled everything from hurricane predictions, engineering automobiles and airplanes, to the discovery of new fundamental physics. More recently, machine learning and in particular deep learning are being used to effectively automate “model” construction and use, dispensing with and/or automating much of the traditional processes in mathematical modeling that establishing model quality and fitness in favor of rapid formulation and prediction. Such customized and implicit models often deliver great results with surprising accuracy and much lower computational costs, especially in a curated (or ground truthed) data rich environment. However, these models do not easily deliver the robustness and reliabilityof traditional mathematical models that have been analyzed for stability, well-posedness and approximation quality. Much research in machine (and in particular deep) learning models is consequently focused on making them more stable and robust and to preventing catastrophic failure. This panel will explore the idea that the process of model analysis for stability, reliability and approximability established for classical mathematical models can inform advances in machine learning methodology to make them more reliable and trustworthy.

This panel will draw parallels to the analysis of traditional mathematical models and illustrate potential research pathways in machine learning. We will use a case study approach with distinguished panelists from traditional machine deep learning and mathematical modeling domains to explore the complementarity attributes of these domains, and the opportunities (and challenges) in bringing together these currently disparate communities and ideas.

Panelists: