The majority of theoretical work in machine learning is done under the assumption of exchangeability: essentially, it is assumed that the examples are generated from the same prob...
Vladimir Vovk, Ilia Nouretdinov, Alexander Gammerm...
—Biasing properly the hypothesis space of a learner has been shown to improve generalization performance. Methods for achieving this goal have been proposed, that range from desi...
The Building-Block Hypothesis appeals to the notion of problem decomposition and the assembly of solutions from sub-solutions. Accordingly, there have been many varieties of GA tes...
Richard A. Watson, Gregory Hornby, Jordan B. Polla...
In this paper we introduce a dynamical model for wireless sensor networks. We obtain a convergent martingale for the broadcast process in such networks. To our knowledge, such mart...
We introduce a kernel-based method for change-point analysis within a sequence of temporal observations. Change-point analysis of an unlabelled sample of observations consists in,...