Kickoff Meeting, Paris, 16-17 May 2018
- 16 May:
- 9h45: The HPC-BigData Challenge: overview. Bruno Raffin
- 10h45: HPC Infrastructures for AI. Olivier Richard & Benoit Pelletier
- 11h: Feedback on using CINES machine. Alexis Joly.
- 11h15: The Gricad/G5K HPDA machine for HPC/BigData processing. Pierre Neyron.
- 11h30: Discussions
- 12h: Talk from ESI-group. Argiris Kamoulakos
- 12h20: (PhD IPL 2018) Bridging in situ/in transit processing with Big Data analytics. Gabriel Antoniu & Patrick Valduriez
- 12h40: (PhD IPL 2018) Scheduling Strategies for High Performance Deep Learning. Olivier Beaumont & Alexis Joly
- 13h30 Lunch
- 15h IPL) Parallel Scikit-Learn with Dask/StarPU. Samuel Thibault & Gael Varoquaux
- 15h20: (PhD IPL 2019) Analysis of Massive Molecular Trajectories through Deep Learning. Guillaume Charpiat & Jérome Henin
- 15h40: (Postdoc IPL 2019) High Performance Deep Reinforcement Learning. Bruno Raffin & Philippe Preux.
- 16h: Algorithmic Challenges: Linear Algebra and Tensors. Olivier Coulaud & Francis Bach
- 16h20: Automatic Differentiation, Backpropagation and Adjoint. Guillaume Aupy
- 17 May: Free discussions and sub-group meetings.
Plenary Meeting, 17 & 18 October 2018, Lyon
- 17 October
- 9h30-10h15: Coffee
- 10h15:
- State of Progress, Bruno Raffin
- Didactic presentation for the HPC community of data integration and BL/DL crossover. Gael Varoquaux
- Deep or Shallow Learning: use cases. Guillaume Charpia
- 12h30: lunch
- 14h: Sub-group Discussions
- 16h: Coffee
- 16h30:
- Use cases / Middleware for data management/ FPGAs. Stéphane Pralet.
- PlantNet experiments on HPC platforms. Alexis Joly
- 17h30: Open Discussions
- Jeudi 18 Octobre:
- 8h30: Sub-group Discussions
- 9h30: State of the Art on AI for Molecular Dynammics.Jérôme Hénin.
- 10h Coffee
- 10h30:
- Compute Infrastructures for HPC-AI. Pierre Neyron
- Parallel polyglot query processing on heterogeneous could data stores with leanxale. Boyan Kolev
- HPC Storage. Emmanuel Jeannot
- 12h30: Lunch
- 14h: Open Discussion
- 16h: Coffee
- 16h30: Summary.
Plenary Meeting, 2-3 April 2019, Paris
- 2 April
- 9h: Welcome Coffee
- 10h: News. Bruno Raffin
- 10h30: Talk: Decentralized Data Processing for the PlantNet Workflow. Pedro Silva (KerData)
- 11h: Discussion: NVRAM impact on BigData and AI. Emmanuel Jeannot
- 11h30: Coffee
- 12h Talk: Decreasing memory consumption while training neural networks with the help of
checkpointing strategies. Alena Shilova - 12h30: Discussion: Smart data center & supercomputer: how to use ML for managing machine resources. Stéphane Pralet
- 15h15: Talk: Asynchronous Gradient Descent. Hadrien Hendrikx (SIERRA)
- 15h45: Discussion: How to leverage the probabilist nature of learning algorithms for asynchronism. Francis Bach
- 16h15: Coffee
- 16h45: Talk: Scikit-Learn – Dask & StarPU. Pierre Glaser
- 17h15: Day Summary. Bruno Raffin
- 3 April
- 9h: Coffee
- 9h15: Talk: Deep Learning for Molecular Dynamics.Loris Felardos
- 9h45: Feedback on the Jean-Zay machine from GENCI. Guillaume Charpiat, Bruno Raffin, Alexis Joly
- 10h: Talk: On-line Learning for Meta-models. Bruno Raffin
- 10h30: Coffee
- 11h: Discussion: Numerical Simulation and AI. Guillaume Charpiat
- 11h30: Discussion: Data and model parallelism for DL. Olivier Beaumont
- 12h15: Lunch
- 14h15: Talk: Deep Reinforcement Learning at Scale: early experiments. Yannis Flet-Berliac (Sequel)
- 14h45: Discussion: Applications of Reinforcement Learning. Philippe Preux
- 15h15: Conv’2019 Workshop (organization, program)
- 15h30: Coffee
- 16h: Summary and Closing Remarks.
Conv’2019, EDF Lab Paris-Saclay – 6 and 7 of November 2019
- We organized with EDF Lab the HPC – AI – BigData Convergence Days (Conv’2019). Go to the program and slides.
Plenary Meeting, 22-23 January 2020, Paris
- 22/01
- 9:15 Coffee
- 9:45 Opening and Activity Report. Bruno Raffin
- 10:30 Overview – MILA perspective on the HPC/IA convergence. Sacha Leprêtre (MILA)
- 10:45 Spike Networks. Loris Felardos
- 11:30 Physics-Inspired Neural Networks. Alessandro Bucci
- 12:15 Deep Reinforcement Learning Infrastructures. Essam Morsi
- 13:30 Lunch
- 15:00 Open discussion: the Jean Zay machine (lead: Loic Esteve)
- Each Jean Zay user should sort-out his/her experience to bring elements to the discussion
- Idris representative will be present (Sylvie Therond, Myriam Peyrounette, Pierre-Francois Lavallée)
- 16:00 Expectations for a day-to-day productive resarch environment, Loic Esteve
- 16:30 IPL: Next actions (lead: Bruno Raffin)
- 17:00:Free group discussions
- 23/01
- 9:00 Cloud and Edge Intelligent Analytics. Daniel Rosendo
- 9:45 Molecular Dynamics & Deep Learning. Jérôme Hénin
- 10:30 Coffee break
- 11:00 Scikit Learn. Olivier Grisel
- 11:30 Open Discussion – Trends
- Suggested topics: auto-deep learning, physics-inspired NN, differential programming.
- 12:00 Keynote – MILA perspective on the HPC/IA convergence. Sacha Leprêtre
- 13:30 Lunch break
- 15:30 MILA Technical Talk – Frederic Costera (visio from MILA)
- 16:00 GENCI. Stéphane Requena
- 16:30 Discussion on potential collaborations INRIA/MILA and Wrap-up.
Plenary Meeting, 30-31 May 2022, Paris
- Monday 30th
- 9:30-10:00 Morning Coffee
- 10:00-10:15 Activity Summary – Bruno Raffin
- 10:15-11:00 External talk: DeepHyper. Romain Negele
- 11:00:-12:30 (20 minutes each)
- Rotor – Lionel Eyraud-Dubois
- Model Based RL – Alena Shilova
- AI on Jean-Zay – Loic Esteve
- 13:30 Lunch
- 15:00-15:45 – External Talk: Hugging Face – Lucile Saulnier – Thomas Wang
- 15:45 Break
- 16:15-17:15 (20 minutes each)
- Learning probability distributions for molecular dynamics – Jerome Henin
- AI in the MD community – Marc Baaden
- Ray – Bruno Raffin
- 17:15: Discussion PEPR IA – PEPR NUMPEX / EU calls / projects
- Tuesday 31th
- 8:45-9:15 Morning Coffee
- 9:15-10:00 – External Talk: ONNX – Xavier Dupré
- 10:00-10:30 Coffee
- 10:30-12:00 (20 minutes each)
- CPU Offloading for GPU NN training – Xunyi Zhao
- AI for physics – Guillaume Charpiat
- Deep Surrogate for CFD – Lucas Meyer
- E2Clab – Daniel Rosendo
- 12:30 Lunch
- 14:30-15:15 External Talk: SINCLAIR Lab – Nicolas Bousquet
- 15:15-15:30 Short coffee break
- 15:30-16:30 (20 minutes each)
- HPC et Data Science: Inria et LATAM – Patrick Valduriez
- Scikit-learn at Scale – Gael Varoauaux
- Simulation based inference – Sofya Dymchenko
- 16:30 Final discussions