Abstract:Neural network ensemble is a learning paradigm where a collection of neural networks is trained for the same task. In this paper, the relationship between the generalization ability of the neural network ensemble and the correlation of the individual neural networks constituting the ensemble is analyzed in the context of combining neural regression estimators, which reveals that ensembling a selective subset of trained networks is superior to ensembling all the trained networks in some cases. Based on such recognition, an approach named GASEN is proposed. GASEN trains a number of individual neural networks at first. Then it assigns random weights to the individual networks and employs a genetic algorithm to evolve those weights so that they can characterize to some extent the importance of the individual networks in constituting an ensemble. Finally it selects an optimum subset of individual networks based on the evolved weights to make up the ensemble. Experimental results show that, comparing with a popular ensemble approach, i.e., averaging all, and a theoretically optimum selective ensemble approach, i.e. enumerating, GASEN has preferable performance in generating ensembles with strong generalization ability in relatively small computational cost. This paper also analyzes the working mechanism of GASEN from the view of error-ambiguity decomposition, which reveals that GASEN improves generalization ability mainly through reducing the average generalization error of the individual neural networks constituting the ensemble.

A Neural Network Selection Approach of Constructing Ensemble Based on CS Algorithm

Genetic Algorithm based Selective Neural Network Ensemble.

A Fast Selective Approach to Neural Network Ensemble

Constrained Programming Based Selective Neural Network Ensemble Method

Selective Neural Network Ensemble Approach Based on Cultural Algorithm

A Selective Approach to Neural Network Ensemble Based on Clustering Technology

Hybrid Ensemble Construction with Selected Neural Networks.

A Constructive Algorithm for Training Neural Network Ensemble

Clustering Algorithm Based Selective Ensemble

Neural Network Ensemble Based on Rough Sets Reduction and Selective Strategy

Study of Ensemble Algorithm and Its Application

Dynamic Neural Network Ensemble Construction For Classification

Ensembling Neural Networks: Many Could Be Better Than All

A Survey of Selective Ensemble Learning Algorithms

A New Ensemble Feature Selection and Its Application to Pattern Classification

Ensemble Algorithm Of Neural Networks And Its Application

Genetic Algorithm Based Selective Ensemble with Multiset Representation

On Dynamic Adaptive Selection Method for Neural Network Ensemble

COMBINING REGRESSION ESTIMATORS: GA-BASED SELECTIVE NEURAL NETWORK ENSEMBLE

An Improved Constructive Neural Network Ensemble Approach To Medical Diagnoses

A Selective GMDH Network Ensemble Algorithm