Julien Herrmann

Contact



Lastname	Herrmann
Firstname	Julien
Email	firstname.lastname[at]inria.fr
Address	Centre de Recherche Inria Bordeaux, 200 Avenue de la Vieille Tour, 33405 Talence

blakasdkasjk
I am currently a Postdoctoral Fellow at the Research Center Inria Bordeaux.
My research interests include scheduling techniques, numerical algorithms for distributed systems, graph algorithms for High Performance Computing, automatic differentiation...
You can find my resume here.

Education and Positions

Postdoc

Laboratory	Research Center Inria Bordeaux
Team	TADaaM
Advisor	Guillaume Aupy, Olivier Beaumont

Postdoc

Laboratory	School of Computational Science and Engineering
School	Georgia Institute of Technology
Advisor	Ümit V. Çatalyürek

PhD

Laboratory	LIP
School	Ecole Normale Supérieure de Lyon
Subject	Algorithms and Scheduling Techniques for Matrix Computations.
Advisors	Yves Robert, Loris Marchal

Master

School	Ecole Normale Supérieure de Lyon
Studies	Computer Science and Applied Mathematics

Publications

International Conferences

• Yusuf Özkaya, Anne Benoit, Bora Uçar, Julien Herrmann, Ümit V. Çatalyürek
  A scalable clustering-based task scheduler for homogeneous processors using DAG partitioning. [pdf]
  IPDPS 2019, Rio de Janeiro, Brasil


• Julien Herrmann, Jonathan Kho, Bora Uçar, Kamer Kaya, Ümit V. Çatalyürek
  Acyclic partitioning of large directed acyclic graphs. [pdf | bibtex]
  CCGRID 2017, Madrid, Spain


• Julien Herrmann, Zachary L. Witter, Nakul Patel, Jonathan Kho, Daniel A. Janies, Ümit V. Çatalyürek
  Visual analytics on the spread of pathogens. [pdf | bibtex]
  BCB 2016, Seattle, Washington, USA


• Henri Casanova, Julien Herrmann, Yves Robert
  Computing the expected makespan of task graphs in the presence of silent errors. [pdf | bibtex]
  ICPP Workshops 2016, Philadelphia, Pennsylvania, USA


• Emmanuel Agullo, Olivier Beaumont, Lionel Eyraud-Dubois, Julien Herrmann, Suraj Kumar, Loris Marchal, Samuel Thibault
  Bridging the Gap between Performance and Bounds of Cholesky Factorization on Heterogeneous Platforms. [pdf | bibtex]
  HCW 2015, Hyderabad, India


• Thomas Hérault, Julien Herrmann, Loris Marchal, Yves Robert
  Determining the optimal redistribution for a given data partition. [pdf | code | bibtex]
  ISPDC 2014, Porquerolles, France


• Mathieu Faverge, Julien Herrmann, Julien Langou, Bradley R. Lowery, Yves Robert, Jack Dongarra
  Designing LU-QR hybrid solvers for performance and stability. [pdf | bibtex]
  IPDPS 2014, Phoenix, Arizona, USA


• Julien Herrmann, Loris Marchal, Yves Robert
  Memory-aware list scheduling for hybrid platforms. [pdf | code | bibtex]
  APDCM 2014, Phoenix, Arizona, USA


• Julien Herrmann, Loris Marchal, Yves Robert
  Model and complexity results for tree traversals on hybrid platforms. [pdf | bibtex]
  Euro-Par 2013 Parallel Processing, Aachen, Germany

Papers

• Julien Herrmann, Yusuf Özkaya, Bora Uçar, Kamer Kaya, Ümit V. Çatalyürek
  Multilevel algorithms for acyclic partitioning of directed acyclic graphs. [pdf | bibtex]
  SIAM Journal on Scientific Computing, Volume 41, 2019


• Henri Casanova, Julien Herrmann, Yves Robert
  Computing the expected makespan of task graphs in the presence of silent errors. [pdf | bibtex]
  Parallel Computing Journal, Volume 75, 2018


• Guillaume Aupy, Julien Herrmann
  Periodicity in optimal hierarchical checkpointing schemes for adjoint computations. [pdf | bibtex]
  Optimization Methods and Software, Volume 32, 2017


• Guillaume Aupy, Julien Herrmann, Paul Hovland, Yves Robert
  Optimal multistage algorithm for adjoint computation. [pdf | bibtex]
  SIAM Journal on Scientific Computing, Volume 38, 2016


• Mathieu Faverge, Julien Herrmann, Julien Langou, Bradley R. Lowery, Yves Robert, Jack Dongarra
  Mixing LU and QR factorization algorithms to design high-performance dense linear algebra solvers. [pdf | bibtex]
  Journal of Parallel and Distributed Computing, Volume 85, 2015


• Julien Herrmann, George Bosilca, Thomas Hérault, Loris Marchal, Yves Robert, Jack Dongarra
  Assessing the cost of redistribution followed by a computational kernel: complexity and performance results. [pdf | code | bibtex]
  Parallel Computing Journal, Volume 52, 2016


• Julien Herrmann, Loris Marchal, Yves Robert
  Memory-aware tree traversals with pre-assigned tasks. [pdf | code | bibtex]
  Journal of Parallel and Distributed Computing (JPDC), Volume 75, 2015


• Marc Baboulin, Jack Dongarra, Julien Herrmann, Stanimire Tomov
  Accelerating Linear System Solutions Using Randomization Techniques. [pdf | bibtex]
  ACM Transactions on Mathematical Software (TOMS), February 2013

Under revision

• Olivier Beaumont, Lionel Eyraud-Dubois, Julien Herrmann, Alexis Joly, Alena Shilova
  Optimal checkpointing strategies for classical networks in pytorch.
  [Under revision in NIPS 2019]


• Olivier Beaumont, Julien Herrmann, Guillaume Pallez (Aupy), Alena Shilova
  Optimal memory-aware backpropagation of deep join networks.
  [Under revision in Philosophical Transactions of the Royal Society]


• Guillaume Aupy, Julien Herrmann
  H-Revolve: A framework for adjoint computation on synchrone hierarchical platforms.
  [Under revision in ACM Transactions on Mathematical Software (TOMS)]

Code



• Disk-Revolve : Library for adjoint computation problems on hierarcical memory systems: here
  
  
  
• Directed Graph Partitioner Library : Coming soon (preliminary version here)
  
  
• Data redistribution followed by a computational kernel: here
  Python code that computes the redistribution that will be performed by the PaRSEC environnement as described in:
  Assessing the cost of redistribution followed by a computational kernel: complexity and performance results.
  and
  Determining the optimal redistribution for a given data partition.
  
  
• Memory aware list scheduling: here
  Python code that runs the memory aware list scheduling on DAGs; Python code that generate the linear program for the exact solution; as described in:
  Memory-aware list scheduling for hybrid platforms.