The balance property in neural network modelling

Abstract

In estimation and prediction theory, considerable attention is paid to the question of having unbiased estimators on a global population level. Recent developments in neural network modelling have mainly focused on accuracy on a granular sample level, and the question of unbiasedness on the population level has almost completely been neglected by that community. We discuss this question within neural network regression models, and we provide methods of receiving unbiased estimators for these models on the global population level.

Keywords:

• Balance property
• unbiased
• binary classification
• logistic regression
• neural network
• classification tree

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 CASdatasets website http://cas.uqam.ca; see also page 55 of the reference manual CASdatasets Package Vignette (2018); we use version 1.0-8 which has been packaged on 2018-05-20.

Additional information

Notes on contributors

Mario V. Wüthrich

Mario V. Wüthrich is Professor in the Department of Mathematics at ETH Zurich, Honorary Visiting Professor at City, University of London (2011-2022), Honorary Professor at University College London (2013-2019), and Adjunct Professor at University of Bologna (2014-2016). He holds a Ph.D. in Mathematics from ETH Zurich (1999). From 2000 to 2005, he held an actuarial position at Winterthur Insurance, Switzerland. He is Actuary SAA (2004), served on the board of the Swiss Association of Actuaries (2006-2018), and is Editor-in-Chief of ASTIN Bulletin (since 2018).