Undercomplete Blind Subspace Deconvolution via Linear Prediction - Statistics > MethodologyReport as inadecuate




Undercomplete Blind Subspace Deconvolution via Linear Prediction - Statistics > Methodology - Download this document for free, or read online. Document in PDF available to download.

Abstract: We present a novel solution technique for the blind subspace deconvolutionBSSD problem, where temporal convolution of multidimensional hiddenindependent components is observed and the task is to uncover the hiddencomponents using the observation only. We carry out this task for theundercomplete case uBSSD: we reduce the original uBSSD task via linearprediction to independent subspace analysis ISA, which we can solve. As ithas been shown recently, applying temporal concatenation can also reduce uBSSDto ISA, but the associated ISA problem can easily become `high dimensional-1. The new reduction method circumvents this dimensionality problem. Weperform detailed studies on the efficiency of the proposed technique by meansof numerical simulations. We have found several advantages: our method canachieve high quality estimations for smaller number of samples and it can copewith deeper temporal convolutions.



Author: Zoltan Szabo, Barnabas Poczos, Andras Lorincz

Source: https://arxiv.org/







Related documents