Singular Learning Theory and Information Criteria

Singular Learning Theory and Information Criteria

A statistical model is called regular if the map from a parameter to a probability density is one-to-one, and if its Fisher information matrix is positive definite. Otherwise, it is called singular. If a statistical model has hierarchical  structures or latent variables, it is not regular but singular. Many statistical models such as neural networks, normal mixtures and matrix factorisations are singular, resulting that the ordinary asymptotic theory does not hold.

In this presentation, we introduce singular learning theory which enables us to clarify the generalisation performance of singular statistical models. Also we show that information teria WAIC and WBIC are constructed. 

Speaker

Sumio Watanabe, Tokyo Institute of Technology

How to attend

If not a member of the Dept. Mathematical Science at the University of Essex, you can register your interest in attending the seminar and request the Zoom’s meeting password by emailing Dr Osama Mahmoud (o.mahmoud@essex.ac.uk).