[Biomathematics] Applications of Random Features in Predicting Epidemics and Function Approximation (Esha Saha)

Speaker: Esha Saha (University of Waterloo)

Talk Title: Applications of Random Features in Predicting Epidemics and Function Approximation

Abstract: With the advent of data-based learning, humans have been trying to develop methods that can learn from data and generalize well based on past information. At the heart of these methods lies the mathematical formulation and analysis of such learning algorithms. One such method which particularly caught attention of researchers recently is the random feature model (RFM). We explore the applications of these class of methods in various tasks. The talk is based on two such applications: (i) short-term epidemic prediction, in which we build a RFM based model paired with delay embeddings to learn and predict the dynamics of an epidemic from incomplete and scarce data; and (ii) high-dimensional function approximation by developing a fast algorithm using a random feature based surrogate model with thresholding algorithms. For both applications we show using numerical and real datasets, that the proposed RFM based methods outperform existing traditional methods.