- Scientific Positionning
Extracting information from huge data sets while leveraging scarce prior information, processing noisy data distributed over a network, making optimal decisions in face of uncertainty — these are typical examples of challenging data science problems considered by the team. The research of DAO is focused on mathematical methods for data science capitalizing on the interplay between
1. mathematical optimization, i.e., the mathematics of doing better with limited resources
2. machine learning, i.e., the tools to undertand/control/modelize of complex systems from data.
The contributions of the members of the team on these themes range from theoretical analysis of problems and algorithms to their numerical applications on real-life data. Our positioning is at the intersection of these three aspects: theory, algorithms, and applications; and and our main research goal is to make a bridge between theory and algorithms, and between algorithms and data science applications.
- Main Research Themes
Machine learning, statistics, mathematical optimization and their interplay make up a vast, active field of research; the DAO team primarily focuses on the following aspects :
– Convex optimization in particular semidefinite and polynomial optimization
– Stochastic optimization robust and chance-constrained optimization
– Distributed, incremental, and randomized numerical algorithms
– Robust, scalable pattern recognition and machine learning
– Applications to academic and real-life problems in computer vision, electricity generation, signal/image processing, and finance.
- Teaching Activities
Members of DAO teach courses related to maths for data science in Grenoble masters (computer science, applied maths, operational research). We actively take part in the evolution of these masters, with the goal to make them match the current demand of data scientists in French and European companies. The team is also a natural environment for students of the “Data Science” major of the Master MSIAM willing to do a PhD.