Christophe Kervazo

 

Christophe Kervazo

PHD STUDENT

Contact Information
E-mail: christophe.kervazo [at] cea [dot] fr
Phone: +33 06 45 72 70 78
Office: 278
Affiliation: IRFU/DEDIP
Supervisor:  Jerome Bobin


Since I finished my PhD, I am not in Cosmostat anymore and this page is not up-to-date. But you can find my new webpage here



Research Interests

My PhD subject is "Learning and image restauration". So far, I have been mainly working on the optimization framework used in sparse linear Blind Source Separation (BSS):

  • I have worked on the extension of GMCA, a sparse BSS algorithm, to large-scale data. I have mainly focused on the case of a large number of sources. The proposed approach consists in using block coordinates methods with intermediate block sizes. I have also participated to the supervision of an internship to tackle the case of a high number of observations / of samples by using mini-batch methods.
  • The second part of my PhD has been devoted to combining heuristic approaches with a sparse BSS algorithm exhibiting interesting mathematical guarantees, namely PALM. The work performed enables a simple hyper-parameter choice, which is shown to be of uttermost importance when performing sparse BSS by minimizing with PALM a classical cost function using a l1 sparse regularization.

In the remaining part of my PhD, I will focus on performing sparse non-linear BSS.

More generally, my main interests are: signal / image processing, blind source separation, optimization, machine learning, sparsity...

 

Teaching

Previously to PhD:

Teaching assistant for the course Probabilities and Statistics for Electrical and Computer Engineering in Georgia Institute of Technology (Atlanta).

During PhD:

1st year:

  • Teacher for the final project in IUT d'Orsay - computer sciences department (20h). Project about object detection and recognition in images.
  • Teacher for the course IN 104 in ENSTA Paristech (20h)

2nd year:

  • Introduction to programming (laboratory work) in IUT d'Orsay - physical measurement department (30h). Practical work in Python.
  • Signal processing laboratory work in CentraleSupélec (48h)
  • Teacher for the final project in IUT d'Orsay - computer sciences department (20h). Project about object detection and recognition in images.

 

Conferences and workshops

Education

I graduated from Supélec in 2015, and obtained a MSc in Electrical Engineering at Georgia Institute of Technology (Atlanta, USA).