Design of ensemble neural network using the Akaike information criterion

Zhao, Zhiye; Zhang, Yun; Liao, Hongjian

doi:10.1016/j.engappai.2008.02.007

Cited by 41 publications

(24 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The AIC indicator has been used by researchers for different aims, e.g. to reduce the subjectivity of the choice between one architecture or another [26], determine the number of hidden neurons [55], and even design committees of networks [75]. The MDL indicator has enabled the reduction of the complexity of different architectures, minimizing the number of weights in domains with scarce data [36], and determining the number of neurons [74] and hidden layers [73] for the optimal model.…”

Section: Stage 4: Quality Measures Of Neural Modelsmentioning

confidence: 99%

Dealing with limited data in ballistic impact scenarios: an empirical comparison of different neural network approaches

et al. 2009

View full text Add to dashboard Cite

In the domain of high-speed impact between solids, the simulation of one trial entails the use of large resources and an elevated computational cost. The objective of this research is to find the best neural network associated with a new problem of ballistic impact, maximizing the quantity of trials available and simplifying their architecture. To achieve this goal, this paper proposes a tuning performance process based on four stages. These stages include existing statistical techniques, a combination of proposals to improve the performance and analyze the influence of each variable. To measure the quality of the different networks, two criteria based on information theory have been incorporated to reflect the fit of the data with respect to their complexity. The results obtained show that the application of an integrated tuning process in this domain permits improvement in the performance and efficiency of a neural network in comparison with different machine learning alternatives

show abstract

Section: Stage 4: Quality Measures Of Neural Modelsmentioning

confidence: 99%

Dealing with limited data in ballistic impact scenarios: an empirical comparison of different neural network approaches

et al. 2009

View full text Add to dashboard Cite

show abstract

“…Los criterios y valores usados se han elegido o fijado en función de las pautas dictadas por Lefebvre et al [60] y su aplicación en problemas prácticos, como la optimización de sistemas eléctricos [61]. Los valores empleados para la configuración del AG se muestran en la Tabla II.…”

Section: ) Fase 3 Arquitectura De Redunclassified

“…El indicador AIC ha sido utilizado por los investigadores con diferentes finalidades, como reducir la subjetividad de la elección entre una arquitectura u otra [66], determinar el número de neuronas ocultas [67] e incluso diseñar comités de redes [61]. Por su parte el indicador MDL ha permitido reducir la complejidad de diferentes arquitecturas minimizando el número de pesos [68], y determinar el número de neuronas [70] y capas ocultas [69] para el modelo óptimo.…”

Section: ) Fase 5 Métricas De Calidadunclassified

Methodology for software development estimation optimization based on neural networks

García¹,

González²,

Colomo‐Palacios³

et al. 2011

IEEE Latin Am. Trans.

View full text Add to dashboard Cite

Abstract-One of the most important tasks of a software development project manager is to produce accurate time and effort estimations. Improving the estimation accuracy is a widely recognized benefit for several software development processes. In order to achieve these objectives, there are proposals based on Artificial Intelligence techniques and specifically artificial neural networks. This paper proposes an optimization methodology for searching the best neural model applicable to the effort estimation of software projects. This will draw on a set of known factors in the early stages of development, outside the complex calculation of function points, which would entail a high level of maturity and definition of the project. This methodology has allowed, on the one hand, ensure the adequacy of the proposed neural network model and, on the other hand, optimize the performance, both in time and accuracy.

show abstract

“…(b) Changing the individual neural network topology structure: to generate individuals and train these network individuals mainly by changing the number of hidden layer nodes [12,13]. The differences of the network topology structure lead to comparatively huge differences of the whole training model, thus generating individuals with differences.…”

Section: Changing Of the Network Characteristicsmentioning

confidence: 99%

A Novel Dynamic Weight Neural Network Ensemble Model

LiKewen

LiuWenying

ZhaoKang

et al. 2015

International Journal of Distributed Sensor Networks

View full text Add to dashboard Cite

Neural network is easy to fall into the minimum and overfitting in the application. The paper proposes a novel dynamic weight neural network ensemble model (DW-NNE). The Bagging algorithm generates certain neural network individuals which then are selected by the -means clustering algorithm. In order to solve the problem that -value cannot be selected automatically in the -means clustering algorithm when conducting the selection of individuals, the -value optimization algorithm based on distance cost function is put forward to find the optimal -values. In addition, for the integrated output problems, the paper proposes a dynamic weight model which is based on fuzzy neural network with accordance to the ideas of dynamic weight. The experimental results show that the integrated approach can achieve better prediction accuracy compared to the traditional single model and neural network ensemble model.

show abstract

Design of ensemble neural network using the Akaike information criterion

Cited by 41 publications

References 30 publications

Dealing with limited data in ballistic impact scenarios: an empirical comparison of different neural network approaches

Dealing with limited data in ballistic impact scenarios: an empirical comparison of different neural network approaches

Methodology for software development estimation optimization based on neural networks

A Novel Dynamic Weight Neural Network Ensemble Model

Contact Info

Product

Resources

About