In the era of data science, it is more than ever necessary to address several challenges by Machine Learning techniques. The assessment and development of new unsupervised, semi-supervised or supervised learning methods, and sophisticated visualizations are needed.

The team MLDS at LIPADE has a long experience in this domain. It has all the necessary skills in applied mathematics and computer science to address different goals in MLDS with various approaches such as Mixture models, Factorization, graph Networks, data-visual analytics and ensemble methods. The researches of the team are realized in theoretical and applied contexts including bioinformatics, text-mining, recommender systems and computer vision. Mathematical models and algorithms are proposed to efficiently reach the objective of knowledge discovery from large and high-dimensional data.

Publications

Softwares

Coclust: a Python package for co-clustering

Coclust provides both a Python package which implements several diagonal and non-diagonal co-clustering algorithms, and a ready to use script to perform co-clustering.

Co-clustering (also known as biclustering), is an important extension of cluster analysis since it allows to simultaneously groups objects and features in a matrix, resulting in both row and column clusters.

Useful Links: Documentation, API, Python Package Index, Source Code, Download

Admin: François Role

Contact: francois.role@parisdescartes.fr

Contributors: Melissa Ailem, Stanislas Morbieu, Mohamed Nadif and François Role

Implemented Co-clustering Algorithms

  • CoclustMod: Ailem, Melissa, François Role, and Mohamed Nadif. Graph modularity maximization as an effective method for co-clustering text data. Knowledge-Based Systems 109 (2016): 160-173.

  • CoclustSpecMod: Labiod, Lazhar, and Mohamed Nadif. Co-clustering for binary and categorical data with maximum modularity. IEEE International Conference on Data Mining (ICDM), 2011. p. 1140-1145

  • CoclustInfo: Govaert, Gérard, and Mohamed Nadif. Mutual information, phi-squared and model-based co-clustering for contingency tables. Advances in Data Analysis and Classification (2016): 1-34

Implemented Clustering Algorithms

  • Spherical k-means: Dhillon, Inderjit S., and Dharmendra S. Modha. Concept decompositions for large sparse text data using clustering. Machine learning 42.1-2 (2001): 143-175.

Seminars

Quantification des effets directs et indirects via l'analyse de médiation

Speaker: Séverine Affeldt, University of Paris Descartes

Time: February 23, 2017, 11:00AM - 12:00AM

Location: Salle de Réunion, Espace Turing, 7 floor, 45 Rue des Saints-Pères, 75006, Paris, France

Outils de clustering diachronique pour analyser l'évolution de la production scientifique

Speaker: Nicolas DUGUE, LORIA

Time: March 10, 2016, 10:00AM - 11:00AM

Location: Salle de Réunion, Espace Turing, 7 floor, 45 Rue des Saints-Pères, 75006, Paris, France

Network reconstruction and causal inference from large scale data

Speaker: Séverine Affeldt, ICAN - Institut de Cardiométabolisme et Nutrition

Time: February 18, 2016, 10:00AM - 11:00AM

Location: Salle de Réunion, Espace Turing, 7 floor, 45 Rue des Saints-Pères, 75006, Paris, France

The Approximation of the Maximal alpha-Consensus Local Community detection problem in Complex Networks

Speaker: Patricia Conde Céspedes, L2TI - University of Paris 13

Time: February 4, 2016, 10:00AM - 11:00AM

Location: Salle de Réunion, Espace Turing, 7 floor, 45 Rue des Saints-Pères, 75006, Paris, France

Fuzzy Clustering

Speaker: Francisco Carvalho, Federal University of Pernambuco

Time: January 18, 2016, 10:30AM - 11:30AM

Location: Salle de Réunion, Espace Turing, 7 floor, 45 Rue des Saints-Pères, 75006, Paris, France





International Master of Machine Learning for Data Science

In the era of data science, people with advanced skills in many areas of Machine Learning and advanced Data Analysis are in high demand. The Master II "Machine Learning for Data Science" at the University of Paris Descartes dates back to 2010 and is under the direction of Pr. Mohamed Nadif since its creation. The master aims at teaching graduate students a wide spectrum of unsupervised, semi-supervised and supervised learning as well as visualization methods. These techniques are commonly used in the context of data science and knowledge discovery for mining and analyzing large and high dimensional data sets. The data sets can come from as various fields as genomics, text analysis, web usage mining, marketing, image and speech processing. Students are also introduced to the use of probabilistic models such as the finite mixtures models, which are important tools in machine learning.

This master II recruits students from computer science and applied mathematics masters. The selection of students is done on programming, statistics and data mining. More than a hundred applications are received every year.

Management

Director: Pr. Mohamed Nadif

Co-Director: Ass. Pr. Lazhar Labiod

Courses, Teachers in Master II and Organization

Common to Machine Learning and Agent-based Computing (courses 76h or 114 h tutorial classes)

  • Decision Theory

  • Agent-Oriented Learning

  • English/French classes for non french speakers

  • Free UE

Specific-Track for Machine Learning courses and Teachers (courses 117 h or 175.5 h tutorial classes)

  • Supervised Learning

  • Unsupervised Learning

  • Finite Mixture Models

  • Co-clustering

  • Data science

  • Visual Analytics

Softwares and Programming languages

  • R

  • SAS

  • Weka

  • Azure ML

  • Python

Team MLDS and teachers

The MLDS team at LIPADE has a long experience in Machine Learning. Members of the team have the necessary skills in applied mathematics and computer science to address different goals in MLDS with various approaches such as mixture models, factorization, graph and networks analysis, data-visual analytics, ensemble methods, etc. The researches of the team are carried out in both theoretical and applied contexts, including bioinformatics, text-mining, recommender systems and computer vision. Mathematical models and algorithms are proposed to efficiently reach the objective of knowledge discovery from large and high dimensional data. The teaching team includes specialists in machine learning with PHDs in the field. They also have extensive experience with the teaching of statistical and data mining techniques.

  • Lazhar Labiod (Paris Descartes University)

  • Mohamed Nadif (Paris Descartes University)

  • François Role (Paris Descartes University)

  • Nicoleta Rogovschi (Paris Descartes University)

  • Allou Samé (IFSTTAR)