We present a tree-structured architecture for supervised learning. The statistical model underlying the architecture is a hi-erarchical mixture model in which both the mixture coefficients and the mixture components are generalized linear models (GLIM’s). Learning is treated as a max-imum likelihood problem; in particular, we present an Expectation-Maximization (EM) algorithm for adjusting the parame-ters of the architecture. We also develop an on-line learning algorithm in which the pa-rameters are updated incrementally. Com-parative simulation results are presented in the robot dynamics domain. 1

Ways of obtaining approximate Bayes factors for generalized linear models are described, based on the Laplace method for integrals. I propose a new approximation which uses only the output of standard computer programs such as GUM; this appears to be quite accurate. A reference set of proper priors is suggested, both to represent the situation where there is not much prior information, and to assess the sensitivity of the results to the prior distribution. The methods can be used when the dispersion parameter is unknown, when there is overdispersion, to compare link functions, and to compare error distributions and variance functions. The methods can be used to implement the Bayesian approach to accounting for model uncertainty. I describe an application to inference about relative risks in the presence of control factors where model uncertainty is large and important. Software to implement the

This paper presents an interactive technique for moving through an immersive virtual environment (or "virtual reality"). The technique is suitable for applications where locomotion is restricted to ground level. The technique is derived from the idea that presence in virtual environments may be enhanced the stronger the match between proprioceptive information from human body movements, and sensory feedback from the computer generated displays. The technique is an attempt to simulate body movements associated with walking. The participant "walks in place" to move through the virtual environment across distances greater than the physical limitations imposed by the electro-magnetic tracking devices. A neural network is used to analyse the stream of coordinates from the head-mounted display, to determine whether or not the participant is walking on the spot. Whenever it determines the walking behaviour, the participant is moved through virtual space in the direction of gaze. We discuss tw...

Motivation. How to selecting a small subset out of the thousands of genes in microarray data is important for accurate classification of phenotypes. Widely used methods typically rank genes according to their differential expressions among phenotypes and pick the top-ranked genes. We observe that feature sets so obtained have certain redundancy and study methods to minimize it. Results. We propose a minimum redundancy – maximum relevance (MRMR) feature selection framework. Genes selected via MRMR provide a more balanced coverage of the space and capture broader characteristics of phenotypes. They lead to significantly improved class predictions in extensive experiments on 5 gene expression data sets: NCI,

This article surveys probabilistic approaches to modeling information retrieval. The basic concepts of probabilistic approaches to information retrieval are outlined and the principles and assumptions upon which the approaches are based are presented. The various models proposed in the development of IR are described, classified, and compared using a common formalism. New approaches that constitute the basis of future research are described

A study by Slater, et al., [1995] indicated that naive subjects in an immersive virtual environment experience a higher subjective sense of presence when they locomote by walking-in-place (virtual walking) than when they push-button-fly (along the floor plane). We replicated their study, adding real walking as a third condition. Our study confirmed their findings. We also found that real walking is significantly better than both virtual walking and flying in ease (simplicity, straightforwardness, naturalness) as a mode of locomotion. The greatest difference in subjective presence was between flyers and both kinds of walkers. In addition, subjective presence was higher for real walkers than virtual walkers, but the difference was statistically significant only in some models. Follow-on studies show virtual walking can be substantially improved by detecting footfalls with a head accelerometer. As in the Slater study, subjective presence significantly correlated with subjects' degree of...

Logistic regression is a workhorse of statistics and is closely related to methods used in Machine Learning, including the Perceptron and the Support Vector Machine. This note compares eight different algorithms for computing the maximum a-posteriori parameter estimate. A full derivation of each algorithm is given. In particular, a new derivation of Iterative Scaling is given which applies more generally than the conventional one. A new derivation is also given for the Modified Iterative Scaling algorithm of Collins et al. (2002). Most of the algorithms operate in the primal space, but can also work in dual space. All algorithms are compared in terms of computational complexity by experiments on large data sets. The fastest algorithms turn out to be conjugate gradient ascent and quasi-Newton algorithms, which far outstrip Iterative Scaling and its variants. 1

Abstract. This paper describes an experiment where the effect of dynamic shadows in an immersive virtual environment is measured with respect to spatial perception and presence. Eight subjects were given tasks to do in a virtual environment. Each subject carried out five experimental trials, and the extent of dynamic shadow phenomena varied between the trials. Two measurements of presence were used- a subjective one based on a questionnaire, and a more objective behavioural measure. The experiment was inconclusive with respect to the effect of shadows on depth perception. However, the experiment suggests that for visually dominant subjects, the greater the extent of shadow phenomena in the virtual environment, the greater the sense of presence.

"Well then, what about the actual getting of wisdom? Is the body in the way or not...? I mean, for example, is there any truth for men in their sight and hearing? Or as poets are forever dinning into our

This paper describes an experiment to assess the influence of immersion on performance in immersive virtual environments. The task involved Tri-Dimensional Chess, and required subjects to reproduce on a real chess board the state of the board learned from a sequence of moves witnessed in a virtual environment. Twenty four subjects were allocated to a factorial design consisting of two levels of immersion (exocentric screen based, and egocentric HMD based), and two kinds of environment (plain and realistic. The results suggest that egocentric subjects performed better than exocentric, and those in the more realistic environment performed better than those in the less realistic environment. Previous knowledge of chess, and amount of virtual practice were also significant, and may be considered as control variables to equalise these factors amongst the subjects. Other things being equal, males remembered the moves better than females, although female performance improved with higher spati...