Splitting methods involve systems of particles that move around in space and time and have the possibility to die out or split into multiple particles at certain times. Such systems can be used to solve complicated estimation and optimization problems. This project aims to (1) create new and faster splitting algorithms; (2) improve the implementation of splitting algorithms, e.g., via distributed computing; (3) find new applications for optimization and estimation that are well-suited to be solved by splitting methods.
The project will deliver new theory and methods for fast and robust statistical learning, inference, and parameter estimation in various application fields such as risk management, decision making, health-care, manufacturing, financial engineering, and system reliability. By providing high-quality solutions that are out of reach of the current state of the art, the project will have large-scale impact on operational efficiency of real-life applications in both scientific and industrial domains.
Invited talks, refereed proceedings and other conference outputs