Neuroscience 2006
February 15-17 Gainesville, FL
Fuzzy Hyperplane Clustering Algorithm and Applications to Sparse
Representations
Pando Georgiev and Anca Ralescu
ECECS Department
University of Cincinnati
E-mail: fpgeorgie,aralescug@ececs.uc.edu
We develop a fuzzy hyperplane clustering algorithm based on finding a skeleton (a union
of hyper-planes) of a given set of data points. The resulting algorithm has a direct
application to the following sparse representation problem: represent a given data matrix
X  mN in the form X = AS, where A  mN, S  mN, as a priori is known that
such a representation exists and the matrix S is sparse in the sense that each column of
it has at least n-m+1 zeros. Applications to fMRI data analysis are considered.