Robust neural learning from unbalanced data samples

Published 27 November 2002

Yi Lu, Hong Guo, L.A. Feldkamp

DOI10.1109/ijcnn.1998.687133

Citations36

TL;DR

The study was conducted on three different neural network architectures, multilayered back propagation, radial basis function, and fuzzy ARTMAP with training methods including duplicating minority class samples and the Snowball technique to solve the classification problem in which the data is unbalanced and noisy.

Abstract

This paper describes the result of our study on neural learning to solve the classification problem in which the data is unbalanced and noisy. Our study was conducted on three different neural network architectures, multilayered back propagation, radial basis function, and fuzzy ARTMAP with training methods including duplicating minority class samples and the Snowball technique. Three major issues are addressed: neural learning from unbalanced data samples, neural learning from noise data, and making intentional biased decisions. The application considered in this study is classifying good(pass)/bad(fail) vehicles. Experiments are conducted on data samples downloaded directly from test sites of automobile assembly.

Keywords

Computer Science

The MIT Press eBooksParallel Distributed Processing

15,227 Citations1986David E. Rumelhart, James L. McClelland +1 more

Neural ComputationFast Learning in Networks of Locally-Tuned Processing Units

4,219 Citations1989John Moody, Christian J. Darken

This work proposes a network architecture which uses a single internal layer of locally-tuned processing units to learn both classification tasks and real-valued function approximations (Moody and Darken 1988).

Choice Reviews OnlineNeural networks and fuzzy systems: a dynamical systems approach to machine intelligence

2,336 Citations1991

This work combines neural networks and fuzzy systems, presenting neural networks as trainable dynamical systems and developing mechanisms and principles of adaption, self-organization, convergence and global stability.

Applied OpticsART 2: self-organization of stable category recognition codes for analog input patterns

1,433 Citations1987Gail A. Carpenter, Stephen Grossberg

ART 2, a class of adaptive resonance architectures which rapidly self-organize pattern recognition categories in response to arbitrary sequences of either analog or binary input patterns, is introduced.

IEEE International Conference on Neural NetworksCapabilities of three-layered perceptrons

400 Citations1988Irie, Miyake

A theorem is proved that three-layered perceptrons with an infinite number of computing units can represent arbitrary mapping if the desired mapping and the input-output characteristics of the computing units satisfy some constraints.

Fuzzy ARTMAP: an adaptive resonance architecture for incremental learning of analog maps

39 Citations2003Gail A. Carpenter, Stephen Grossberg +3 more

Fuzzy ARTMAP achieves a synthesis of fuzzy logic and adaptive resonance theory (ART) neural networks and realizes a new minimax learning rule that conjointly minimizes predictive error and maximizes code compression or generalization.

Pattern RecognitionResolving multifont character confusion with neural networks

36 Citations1993Jin Wang, Jack Jean

A set of feedforward neural networks and a rule invocation procedure are added to a conventional character recognition system to help resolve similar character confusion and this approach has the advantage of automating the feature selection and feature extraction processes, and it is more robust to noise.