The team Modeling and Knowledge Discovery develops data mining methods
and algorithms, and has to face new challenging opportunities in the
complex system domain (work in collaboration with the Complex System
Institute IXXI), more particularly in the field of in silico biology.
In this PhD research subject, we propose to develop data mining models
and methods to study dynamic systems, being either artificial (e.g.,
simulation of proteins-proteins interactions) or real (e.g., gene
expression in cells). The first step will be to study analogies between
complete methods for pattern extractions in large Boolean matrices and
the search for regularities in large graphs or collection of graphs.
Dealing with dynamic graphs or with graphs specifying the dynamic of a
system, leads naturally to consider Boolean hypercubes to handle the
temporal dimension and/or features associated to vertices and edges.
Other directions that seem promising include to investigate the
specificities of the graphs that are used (e.g., interaction graphs) and
to take advantage of the properties of the families of patterns
occurring in these graphs (e.g., error tolerant patterns).