• When: 6 and 7 of November 2019
  • Where:  EDF Lab Paris-Saclay
  • Contact:
  • Registrations:  Closed
  • News:
    • 7/11/2019:  More than 200 registrations with a good mix of academic and industrial attendees. Great talks, great discussions, very fritfull event.
    • 14/10/19: The program is now online


High Performance Computing (HPC), Machine Learning and Big Data are supported by distinct communities developing different scientific and technological approaches and targeting different application domains. Today a trend is emerging to converge these domains with high benefits expected from this cross-fertilization.

Deep learning has a strong appetite for compute power, making HPC a natural target. Strong progress has been made to efficiently parallelize the training phase. As an example, ResNet-50 training evolved from 29 hours on 8 GPUS in 2016 to 224 seconds on 2176 GPUs in 2018. Numerical simulations, the first customer of supercomputers, are evolving to integrate machine learning algorithms at various level, for output analysis or get new fast numerical solvers by training for instance. Machine learning can also be used to improve supercomputer management for a better QoS and resource usage.

Conv’2019  is  intended to bring together stakeholders from the academy and industry interested in the various aspects of this emerging convergence between Artificial Intelligence, Big Data and High Performance Computing.

The goal of these 2 days is to create a space for discussions and reflections structured around keynote speeches, scientific presentations, panels. Do not hesitate to propose a presentation or panel.

Topics covered include (not limited):
  •   Converged applications mixing simulations and analytics (e.g., data-enhanced Digital Twins)
  •   Numerical simulations, data analytics and machine learning: coupling and integration
  •   Distributed and parallel machine learning
  •   Extreme-scale learning
  •   Extreme-scale data analytics, High-Performance Data Analytics
  •   Machine learning for optimizing large digital infrastructures (supercomputers, cloud, edge)
  •   Innovative hardware and software architectures and smart administration of digital infrastructures in support of heterogeneous applications (numerical simulations, data analytics, shallow and deep learning)
  •   Streaming, curation, processing and  storage of massive data for learning in converged scenarios
  •   Scalable online learning and analytics


This event is supported by INRIA and EDF R&D.


Comments are closed.