A0734
Title: Mixed moving average field guided learning for spatiotemporal data
Authors: Imma Valentina Curato - TU Chemnitz (Germany) [presenting]
Abstract: Influenced mixed moving average fields are a versatile modeling class for spatiotemporal data. However, their predictive distribution is not generally known. Under this modeling assumption, a novel spatiotemporal embedding and a theory-guided machine learning approach are defined that employs a generalized Bayesian algorithm to make ensemble forecasts. Lipschitz predictors are employed, and fixed-time and any-time PAC Bayesian bounds are determined in the batch learning setting. Performing causal forecast is a highlight of the methodology as its potential application to data with spatial and temporal short and long-range dependence.
