Statistics Seminar: Christopher Yau (University of Manchester)

Abstract: Variational Autoencoders (VAE) have become a popular
approach for implementing Bayesian latent variable models for
applications involving high-dimensional and/or large sample size data
sets in recent years. By leveraging the use of modern programming
frameworks for constructing models based on deep neural networks, VAEs
have been extensively used for unsupervised dimensionality reduction
particularly in image analysis. In this talk, I will describe some
novel variants of VAEs that we have developed to improve their
applicability for problems in computational biology involving ‘omics
datasets.