For high-dimensional, complex data sets, structured dimensionality reduction methods are essential in order to enable useful further processing. While image data are directly accessible and rather easy to interpret, time series data such as audio seem to be harder to understand, cluster and classify. In our work, we apply an approach based on a generalization of time-frequency localization operators (mixed-stated localization operators), which is inspired by quantum information theory in order to get an accesse to the intrinsic structure and effective dinemsionality of given 1D-data sets. This is joint work with Franz Luef and Eirik Skrettingland.
We kindly ask you to use your full name