International Journal on Artificial Intelligence Tools Vol. 16, No. 6 (2007) 967–979 c  World Scientific Publishing Company LARGE SCALE MULTIKERNEL RELEVANCE VECTOR MACHINE FOR OBJECT DETECTION DIMITRIS TZIKAS * , ARISTIDIS LIKAS † and NIKOLAS GALATSANOS ‡ Department of Computer Science, University of Ioannina Ioannina, 45110, Greece * tzikas@cs.uoi.gr † arly@cs.uoi.gr ‡ galatsanos@cs.uoi.gr The Relevance Vector Machine(RVM) is a widely accepted Bayesian model commonly used for regression and classification tasks. In this paper we propose a multikernel version of the RVM and present an alternative inference algorithm based on Fourier domain computation to solve this model for large scale problems, e.g. images. We then apply the proposed method to the object detection problem with promising results. Keywords : Relevance vector machine; object detection; image analysis. 1. Introduction The Relevance Vector Machine (RVM) 1 is a Bayesian treatment of the linear model given by: y(x)= M  i=1 w i φ i (x) , (1) where {φ i (x)} M i=1 is a set of basis functions. Learning on such a model, is the process of estimating the weights {w i } M i=1 , using a training set {(  x n ,t n )} N n=1 . The weights are typically assigned those values that maximize the likelihood of the training set, however the training examples must be significantly more than the parameters in order to achieve good generalization performance. The RVM overcomes this limita- tion by following Bayesian principles and assuming prior knowledge for the model. Specifically, a suitable hierarchical prior distribution is assumed for the weights of the model, which has most probability mass concentrated in sparse solutions, mean- ing that it forces most of the weights to be assigned to zero values. 1 This results in pruning basis functions that are not sufficiently supported by the training data. There are several reasons to seek sparse solutions: • Sparseness automatically adjusts the complexity of the model, thus very complex models may be considered. 967