Points to talk:

  * DMultiMads EB: what does not work, what works, what I can try => is it too close to DMS ? (better convergence proof) but...
  * DMultiMads PB: what is running; do we have to introduce a change according to the first algorithm ?
  * Internship.
  * Heuristic benchmarks workshop.
  * Mathilde Perreyga: folder on epsilon machine: ask the possibility to check it.