Exploratory project Persyvact
Building tools for analyzing high dimensional and structured
data. The tools proposed by Persyvact are based on cutting-edge
mathematical and algorithmic developments.
TIMC-IMAG
Co-PI:M Blum
4 people
gipsa-lab
Co-PI: L Condat
8 people
LJK
Co-PI: M Clausel
9 people
Kick-off meeting Persyvact, 16 people
20 septembre 2013, château de la commanderie
Organization of 8 scientific events
March 31-April 1 Spring school on
machine learning with the EA Khronos
April 2-3 Statlearn 2015 in Grenoble
2014
Seminar on machine learning for personalized genomics
Workshop on graphical models
Astrostatistics in Grenoble
2013
Seminar on Bayesian approaches
Workshop on extremes and copulas
Interdisciplinary meeting on Human Color Perception
TIMC-IMAG
M Blum
O François
gipsa-lab
O Michel
L Condat J Chanussot
LJK
J Mairal
S GirardM Clausel
V Perrier
PhD cosupervision
Machine learning
for genomic imputation
and phasing
Spatially regularized NMF
for genetic application
Modeling of anisotropic
textures with monogenic
wavelets
Sequential data
learning
Data X in genomics
Locus 1 Locus 2 Locus 3
Indiv 1 1 0 2
Indiv 2 0 2 0
Indiv 3 0 0 0
Indiv 4 0 1 1
Indiv 5 1 1 1
• 3 billion base pairs in the human genome
• 1 000- 5 000 Euros for whole genome
sequencing (millions of loci)
Local correlation
Spatialcorrelation
Spatially regularized NMF for genetic application
PhD Kevin Caye (supervision O François, O Michel)
• Clustering (Vincent Delerm) « T'es dans la
catégorie de celles qui lisent Bukowski en
trouvant super naze de mettre les gens dans
des cases. »
• In genetics, we rather estimate admixture
coeffients that correspond to the respective
contributions of the parental populations to
the admixed individuals.
Machine learning for genomic imputation and
phasing
PhD Thomas Dias-Alves (supervision M Blum, J Mairal)
Maternal copy
1100110
Paternal copy
0000100
11002x0
Phasing
Application médicale pour l’étude du déterminisme génétique des
maladies
Haplotypes
Genotypes
Optimization problem
An example of phasing
Perspectives
Project-team Persyvact2 for the ongoing call
“Structured models and algorithmic methods for
the analysis of complex data in high dimension.
Application for health monitoring. ”

Pres blum persyvact_04032015

  • 1.
    Exploratory project Persyvact Buildingtools for analyzing high dimensional and structured data. The tools proposed by Persyvact are based on cutting-edge mathematical and algorithmic developments. TIMC-IMAG Co-PI:M Blum 4 people gipsa-lab Co-PI: L Condat 8 people LJK Co-PI: M Clausel 9 people Kick-off meeting Persyvact, 16 people 20 septembre 2013, château de la commanderie
  • 2.
    Organization of 8scientific events March 31-April 1 Spring school on machine learning with the EA Khronos April 2-3 Statlearn 2015 in Grenoble 2014 Seminar on machine learning for personalized genomics Workshop on graphical models Astrostatistics in Grenoble 2013 Seminar on Bayesian approaches Workshop on extremes and copulas Interdisciplinary meeting on Human Color Perception
  • 3.
    TIMC-IMAG M Blum O François gipsa-lab OMichel L Condat J Chanussot LJK J Mairal S GirardM Clausel V Perrier PhD cosupervision Machine learning for genomic imputation and phasing Spatially regularized NMF for genetic application Modeling of anisotropic textures with monogenic wavelets Sequential data learning
  • 4.
    Data X ingenomics Locus 1 Locus 2 Locus 3 Indiv 1 1 0 2 Indiv 2 0 2 0 Indiv 3 0 0 0 Indiv 4 0 1 1 Indiv 5 1 1 1 • 3 billion base pairs in the human genome • 1 000- 5 000 Euros for whole genome sequencing (millions of loci) Local correlation Spatialcorrelation
  • 5.
    Spatially regularized NMFfor genetic application PhD Kevin Caye (supervision O François, O Michel) • Clustering (Vincent Delerm) « T'es dans la catégorie de celles qui lisent Bukowski en trouvant super naze de mettre les gens dans des cases. » • In genetics, we rather estimate admixture coeffients that correspond to the respective contributions of the parental populations to the admixed individuals.
  • 8.
    Machine learning forgenomic imputation and phasing PhD Thomas Dias-Alves (supervision M Blum, J Mairal) Maternal copy 1100110 Paternal copy 0000100 11002x0 Phasing Application médicale pour l’étude du déterminisme génétique des maladies Haplotypes Genotypes
  • 9.
  • 10.
  • 11.
    Perspectives Project-team Persyvact2 forthe ongoing call “Structured models and algorithmic methods for the analysis of complex data in high dimension. Application for health monitoring. ”