Linear program differentiation for singlechannel speech separation 
Pearlmutter, Barak A.; Olsson, Rasmus K.



Many apparently difficult problems can be solved by reduction to linear programming. Such problems are often subproblems within larger systems. When gradient optimisation of the entire larger system is desired, it is necessary to propagate gradients through the internallyinvoked LP solver. For instance, when an intermediate quantity z is the solution to a linear program involving constraint matrix A, a vector of sensitivities dE/dz will induce sensitivities dE/dA. Here we show how these can be efficiently calculated, when they exist. This allows algorithmic differentiation to be applied to algorithms that invoke linear programming solvers as subroutines, as is common when using sparse representations in signal processing. Here we apply it to gradient optimisation of over complete dictionaries for maximally sparse representations of a speech corpus. The dictionaries are employed in a singlechannel speech separation task, leading to 5 dB and 8 dB targettointerference ratio improvements for samegender and oppositegender mixtures, respectively. Furthermore, the dictionaries are successfully applied to a speaker identification task.

Keyword(s):

Computer Science; Differentiation; Gradient methods; Linear programming; Matrix algebra; Signal representation; Speaker recognition; Constraint matrix; Gradient optimisation; Linear program differentiation; Signal processing; Singlechannel speech separation; Sparse representation; Speaker identification task. 
Publication Date:

2006 
Type:

Conference item 
PeerReviewed:

Yes 
Institution:

Maynooth University 
Citation(s):

Pearlmutter, Barak A. and Olsson, Rasmus K. (2006) Linear program differentiation for singlechannel speech separation. In: Proceedings of the 16th IEEE Signal Processing Society Workshop on Machine Learning for Signal Processing, 6th  8th September, 2006, Arlington, VA . 
File Format(s):

application/pdf 
Related Link(s):

http://mural.maynoothuniversity.ie/1376/1/getPDF3.pdf 
First Indexed:
20200131 06:08:08 Last Updated:
20200402 07:46:48 