Home Research
Home Research

Learning and optimization in large scale scenarios

Research Area: Medical Image Analysis
Status: In progress  
Faculty: Alessandro Verri Participants: Curzio Basso, Matteo Santoro, Alessandra Staglianò

Modern diagnostic imaging produces massive amount of data leading to large-scale problems. Effective solutions to these problems cannot be obtained by simply scaling up the computing resources but require the development of novel computational schemes.

The problems we face range from determining an exact solution (e.g. dense linear systems which do not fit into physical memory), to dealing with redundancy in very large data samples, and developing parallel approaches leveraging on new architecture (GPU, FPGA, multi-core, and grid). We are focusing on the parallelization of iterative projection algorithm and the development of on-line stochastic optimization methods.

The outcome of this effort is beneficial to our entire research lab and is carried out in conjunction with the learning theory group.