The goal of WP1 is to exploit other structural properties of data driven problems to derive algorithms adapted to the modeling assumptions both from the design standpoint and the convergence one. This viewpoint will also allow to go beyond convexity, and successfully solve some nonconvex, but structured, problems.
Principal Investigator: IBSPAN
The goal of WP2 is to leverage the intrinsic acceleration provided by recent multicore computing architectures or GPUs by developing new decomposition strategies that ideally adapt to the problem structure (in the data, in the objective function, and in the computational platform).
Principal Investigator: CentraleSupelec
Stochastic and incremental methods will be the subject of WP3, whose main goal is to apply these methods as low cost regularization procedures to solve huge scale linear inverse problems.
Principal Investigator: UPB
The goal of WP4 is to build bridges between the new advances that will result from WP1-WP3 and what is currently deployed by practitioners using enterprise-ready tools and components. We will adapt and implement the methods studied in WP1-3 to solve problems of interest for TraDE-Opt partners and crucial for the economic growth of Europe in three areas: smart mobility, imaging scanners and devices, transition to smart industries
Principal Investigator: UNI GRAZ