We present a method to recover scenes deteriorated by superposition of transparent and semi-reflected contributions, as appear in reflections off windows. Separating the superimpo...
Abstract. We evaluated methods of protein classification that use kernels built from BLAST output parameters. Protein sequences were represented as vectors of parameters (e.g. simi...
A common approach in structural pattern classification is to define a dissimilarity measure on patterns and apply a distance-based nearest-neighbor classifier. In this paper, we i...
Probability theory is the framework for making decision under uncertainty. In classification, Bayes' rule is used to calculate the probabilities of the classes and it is a bi...
Mohammed J. Islam, Q. M. Jonathan Wu, Majid Ahmadi...
General graph matching methods often suffer from the lack of mathematical structure in the space of graphs. Using kernel functions to evaluate structural graph similarity allows u...