About the research

Due to ever increasing data acquisitio ncapability in important fields such as genomics, remote sensing, finance, and social networks the amount of data is growing in order to extract as much information as possible. For further progress the development of efficient algorithms for processing high dimensional data is crucial. Typically the datasets arising are corrupted with noise and have many more variables than observations which complicates processing. Moreover the data sometimes have known structure such as positivity, sparseness and smoothness. Multivariate signal processing can be viewed as a framework that provides tools for extracting information out of such high dimensional data.

In this project we investigate three important multivariate signal processing data models: Hyperspectral Unmixing, sparse variable Reduced Rank Regression (svRRR), and noisy Independent Component Analysis (nICA). The main objective of the project is to develop signal processing algorithms that are efficient and can extract useful information out of real data sets in remote sensing and genomics.

The main impact of this project is the start of a multivariate signal processing laboratory at the University of Iceland. In Iceland there are strong research teams in fields that require processing of high dimensional data, for example in genomics and remote sensing. What is lacking is a center that focuses primarily on the development on algorithms for this processing.

Keywords: Signal processing, Remote sensing, Genomics

See further details