We propose an extension of the mixture of factor (or independent component) analyzers model to include strongly super-gaussian mixture source densities. This allows greater economy in representation of densities with (multiple) peaked modes or heavy tails than using several Gaussians to represent these features. We derive an EM algorithm to find the maximum likelihood estimate of the model, and...

One of the keystones of the canceled BTeV experiment (proposed at Fermilab’s Tevatron) was its sophisticated threelevel trigger. The trigger was designed to reject 99.9% of lightquark background events and retain a large number of B decays. The BTeV Pixel Detector provided a 3-dimensional, high resolution tracking system to detect B signatures. The Level 1 pixel detector trigger was proposed as...

In this report, we propose a statistical model to deal with the discrete-distribution data varying over time. The proposed model – HMM+DM – extends the Dirichlet mixture model to the dynamic case: Hidden Markov Model with Dirichlet mixture output. Both the inference and parameter estimation procedures are proposed. Experiments on the generated data verify the proposed algorithms. Finally, we di...

1.1 Classification Model Before presenting in more details the Gaussian Mixture Model (GMM) classification process, it is worthwhile to consider what “classification” actually means. According to [3], a “classification model” is made of three main parts : • a transducer : in the case of music this would typically be the A/D conversion chain of the sound. • a feature extractor : it extracts sign...

Gaussian mixture models (GMMs) are a convenient and essential tool for the estimation of probability density functions. Although GMMs are used in many research domains from image processing to machine learning, this statistical mixture modeling is usually complex and further needs to be simplified. In this paper, we present a GMM simplification method based on a hierarchical clustering algorith...

Decision-bound models of categorization like General Recognition Theory (GRT: Ashby & Townsend, 1986) assume that people divide a stimulus space into different response regions, associated with different categorization decisions. These models have traditionally been applied to empirical data using standard model-fitting methods like maximum likelihood estimation. We implement the GRT as a Bayes...

In this paper, we investigate questions involving the impact of a multitude of covariates and their interactions on the scores of a comprehensive exit exam from 121 undergraduate students in Texas State University via a hierarchical Bayesian mixture model. The model uses a mixture of Beta distributions, inflated for the purposes of modeling the behavior of students who are no-shows for the exam...