Towards Black-box Parameter Estimation

Amanda Lenzi (University of Edinburgh)

Friday 9th February 13:00-14:00 JOSEPH BLACK Room A504

Abstract

As soon as we move away from Gaussian processes as the canonical model for dependent data, likelihood computation becomes effectively impossible, and inference is too complicated for traditional estimation methods. Consider, for instance, datasets from finance or climate science, where skewness and jumps are commonly present. In those cases, calculating the likelihood in closed form is often impossible, even with small datasets.

This talk presents deep learning-based procedures to estimate parameters of statistical models for which simulation is easy, but likelihood computation is challenging. These estimators are fast to compute, likelihood-free, and bootstrap-based uncertainty quantification can be easily computed due to their amortized nature. I will demonstrate the applicability of the proposed approaches to quickly and optimally obtain estimates and uncertainty for parameters from non-Gaussian models with complex spatial and temporal dependencies.

Add to your calendar

Download event information as iCalendar file (only this event)