COMON
Pierre
Directeur de recherche CNRS
Research
  • Blind techniques: equalization, identification, source separation (past)
  • Digital communications, interception, surveillance (past)
  • Tensor decompositions
  • Data sciences, Data mining
  • High-order statistics
  • Biomedical (past)
  • Machine learnnig
  • Environment


The Research world publishes too much

  • More publications also means more reviews to perform, or poorer quality papers.
  • Why not fix a limit of 2 journal papers per researcher per year? It would be largely sufficient. The quality is more important than quantity. Becaue of quantity-based evaluation criteria, we are all obliged to undergo this absurd curse of accumulation.
  • More conferences means more energy wasting.
  • The 10 reasons of Donald Geman to abolish conference papers.
  • Greeen benefits of video conferencing.

Research funding through calls for projects is detrimental to it

  • Actually, there is no competent expert without potential conflict of interest. Asking that the applications detail the avenues considered and entrusting the expertise to specialists (often foreign) only offers our assets to the competition.
  • The call for projects process that the State implements is precisely the one described by Léo Szillard in 1948 in his short story "The Mark Gable foundation", a process whose goal was to slow down research!
  • This is what we have been experiencing since the creation of the ANR: (1) researchers help write the call, (2) other researchers write and submit their application, with a 15% chance of success, (3) nominated researchers evaluate the applications and then rank them in order of merit, mostly honestly - but there are conflicts of interest, (4) researchers who have been fortunate enough to get funding are asked to write regular progress reports, (5) other researchers are asked to review the latter progress reports. So many reviews and so much time wasted during which research does not advance!
  • The State has cut support for virology research until 2019. See the edifying article by Bruno Canard published in 2020 (in French).


Source codes

  • Computer (mainly Matlab) codes for tensor decomposition can be found at the Tensor Package page.


European calls

This section has been moved to GIPSA intranet

Grenoble Images Parole Signal Automatique laboratoire

UMR 5216 CNRS - Grenoble INP - Université Joseph Fourier - Université Stendhal