SPARSe: Strategic Planning and Analysis for Reduced Sensing in Inverse Problems

This project rethinks inverse problems by directly inferring quantities of interest instead of full images. Using Bayesian inference, manifold models, and optimal experimental design, it dramatically reduces data and computation requirements, enabling safer medical imaging, efficient seismic analysis, and deployable industrial solutions.
Sparse project logo

Project information

Project duration

-

Funded by

Research Council of Finland - Academy Research Fellow

Project coordinator

University of Oulu

Contact information

Project leader

Other persons

Project description

Many modern technologies, from medical scanners to seismic imaging, work by solving inverse problems: they collect indirect measurements and use mathematics to infer what is happening inside a system. Today, these methods often reconstruct large, detailed images even when only a small piece of information is actually needed, making them slow, costly, and sometimes unsafe.

This project takes a different approach. Instead of reconstructing everything, it focuses directly on the most important information, such as the shape of a tumor or the boundary between geological layers. By doing so, it can achieve accurate results using far fewer measurements.

The project combines advanced mathematics, statistics, and data-driven methods to measure uncertainty and decide where and how data should be collected. This leads to safer medical imaging with less radiation, faster seismic analysis, and efficient tools that scientists and engineers can use in practice.