Sparse Models for the Decomposition of Spectral Data
Frank-Michael Schleif
University of Leipzig

Abstract:

Measurement systems in the life sciences are frequently based on spectral data. Thereby the measured substances are in general a mixture and hence also the measured signals are mixed spectra. The deconvolution of such signal is challenging and different approaches have been proposed to decompose signal by means of blind source separation. Taking the measurement technique into account alternative approaches focusing on non-negative sparse representations can be expected to be more effective. The talk provides an introduction into the field of signal decomposition by different kinds of techniques and gives examples for the analysis of mass spectrometric data.
back to the list of talks