On Wednesday, April 3rd, 2024, the seminar series hosted a talk by Emmanuel Akeweje, PhD Researcher at the School of Computer Science and Statistics in Trinity College Dublin. Emmanuel talked about ensemble clustering techniques for mixtures of Gaussian processes. Details for the talk are below.
Title
Ensemble Clustering for Learning Mixtures of Gaussian Processes
Abstract
We develop an ensemble clustering framework for identifying the latent cluster labels of functional data that are generated from a Gaussian process mixture. Our approach capitalizes on a critical feature of Gaussian random functions: When the functional data are from a Gaussian process mixture, the projection coefficients of the functional data onto any given projection function conform to a univariate Gaussian mixture model (GMM). Therefore, through the execution of multiple one-dimensional projections and the learning of a univariate GMM for each projection, we will obtain an ensemble of GMMs. With each GMM representing a base clustering, we then apply the ensemble clustering technique to derive the consensus clustering. The computational complexity for identifying the cluster labels is much lower than that of state-of-the-art methods, and we provide theoretical guarantees on the identifiability and learnability of Gaussian process mixtures. Extensive experimentation on both synthetic and real datasets validates the superiority of our method over existing techniques.