Multiscale data analysis - information fusion and constant-time clustering

Abstract

We describe the use of the wavelet transform for multivariate data analysis problems. In prediction, a multiscale transform of time-varying data can allow forecasts of each scale, followed by a combining of the individual forecasts. The use of a wavelet transform with noise modeling for point pattern clustering can lead to the result, which initially appears counterintuitive, of clustering in constant computational time, O (1).