Functional Data Analysis
Conference
Category: International Society for Business and Industrial Statistics (ISBIS)
Proposal Description
The session will explore new and active areas in functional data analysis. There will be four talks on recent developments in spatio-temporal and time series of functional data. One talk will be on the estimation of spectral density when the domain of the function is large. They derive a deep learning estimator and prove that it is a universal approximator to the spectral density under general assumptions. The second talk provides a Bayesian method for detection of change points in mean, variance and autocovariance operator in functional time series. Efficient Gibbs sampling, dynamic linear modes and robust Kalman filtering are employed to identify the locations of individual or simultaneous change. The third talk explores nonparametric methods and supervised learning in the context of FDA. Applications will be considered in various domains like humidity patterns in South Africa, medical scans like fMRI images, financial markets including cryptocurrency prices.