Monotone and partially monotone neural networks

H.A.M. Daniëls, M.V. Velikova

Research output: Contribution to journalArticleScientificpeer-review

Abstract

In many classification and prediction problems it is known that the response variable depends on certain explanatory variables. Monotone neural networks can be used as powerful tools to build monotone models with better accuracy and lower variance compared to ordinary nonmonotone models. Monotonicity is usually obtained by putting constraints on the parameters of the network. In this paper, we will clarify some of the theoretical results on monotone neural networks with positive weights, issues that are sometimes misunderstood in the neural network literature. Furthermore, we will generalize some of the results obtained by Sill for the so-called min-max networks to the case of partially monotone problems. The method is illustrated in practical case studies.
Original languageEnglish
Pages (from-to)906-917
JournalIEEE Transactions on Neural Networks
Volume21
Issue number6
Publication statusPublished - 2010

Fingerprint

Neural networks

Cite this

@article{181af53b182a4f23b5fa1014e07cfbac,
title = "Monotone and partially monotone neural networks",
abstract = "In many classification and prediction problems it is known that the response variable depends on certain explanatory variables. Monotone neural networks can be used as powerful tools to build monotone models with better accuracy and lower variance compared to ordinary nonmonotone models. Monotonicity is usually obtained by putting constraints on the parameters of the network. In this paper, we will clarify some of the theoretical results on monotone neural networks with positive weights, issues that are sometimes misunderstood in the neural network literature. Furthermore, we will generalize some of the results obtained by Sill for the so-called min-max networks to the case of partially monotone problems. The method is illustrated in practical case studies.",
author = "H.A.M. Dani{\"e}ls and M.V. Velikova",
year = "2010",
language = "English",
volume = "21",
pages = "906--917",
journal = "IEEE Transactions on Neural Networks",
issn = "1941-0093",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "6",

}

Monotone and partially monotone neural networks. / Daniëls, H.A.M.; Velikova, M.V.

In: IEEE Transactions on Neural Networks, Vol. 21, No. 6, 2010, p. 906-917.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Monotone and partially monotone neural networks

AU - Daniëls, H.A.M.

AU - Velikova, M.V.

PY - 2010

Y1 - 2010

N2 - In many classification and prediction problems it is known that the response variable depends on certain explanatory variables. Monotone neural networks can be used as powerful tools to build monotone models with better accuracy and lower variance compared to ordinary nonmonotone models. Monotonicity is usually obtained by putting constraints on the parameters of the network. In this paper, we will clarify some of the theoretical results on monotone neural networks with positive weights, issues that are sometimes misunderstood in the neural network literature. Furthermore, we will generalize some of the results obtained by Sill for the so-called min-max networks to the case of partially monotone problems. The method is illustrated in practical case studies.

AB - In many classification and prediction problems it is known that the response variable depends on certain explanatory variables. Monotone neural networks can be used as powerful tools to build monotone models with better accuracy and lower variance compared to ordinary nonmonotone models. Monotonicity is usually obtained by putting constraints on the parameters of the network. In this paper, we will clarify some of the theoretical results on monotone neural networks with positive weights, issues that are sometimes misunderstood in the neural network literature. Furthermore, we will generalize some of the results obtained by Sill for the so-called min-max networks to the case of partially monotone problems. The method is illustrated in practical case studies.

M3 - Article

VL - 21

SP - 906

EP - 917

JO - IEEE Transactions on Neural Networks

JF - IEEE Transactions on Neural Networks

SN - 1941-0093

IS - 6

ER -