Speaker
Aidan Mullins
(Newcastle University)
Description
The work of Sir David Cox and Halbert White is dedicated to the poverty of Maximum Likelihood fits under model misspecification, including poor data quality, as well as the necessary remedial, complex regularization. A literature on an alternative to Maximum Likelihood, Maximum Mean Discrepancy (MMD), has developed over the last seven decades. Recent contributions mathematically prove MMD’s ability to handle misspecification without regularization. I present a framework for assessing a MMD-fitted model’s predictive capacity without requiring the separation of data into testing and training sets, with novel contributions found in resample-free configurations of the framework.
Author
Aidan Mullins
(Newcastle University)
Co-authors
Prof.
Chris Oates
(Newcastle University)
Dr
Markus Rau
(Newcastle University)
