Efficient training of RBF neural networks for pattern recognition

Lampariello, F.; Sciandrone, Marco

doi:10.1109/72.950152

Cited by 43 publications

(14 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, linear methods can be applied to build the relationship between the transitional space and output space. This process is the principle of conventional radial basis function networks (RBFN) [17,18]. The advantages of integrating RBF with PLS over conventional RBFN are to make full use of the information of all the samples and to solve the problems of determining the radial bases (such as the local optima of the radial basis vectors obtained through a K-means algorithm).…”

Section: Description Of Rbf-plsmentioning

confidence: 99%

Predicting bulk ambient aerosol compositions from ATOFMS data with ART-2a and multivariate analysis

Zhao

Hopke

Qin

et al. 2005

Analytica Chimica Acta

View full text Add to dashboard Cite

The aerosol time-of-flight mass spectrometry (ATOFMS) has not generally been used to provide a quantitative estimation of chemical compositions of ambient aerosols. In an initial study, the possibility of developing a calibration model to predict chemical compositions from ATOFMS data was demonstrated, but because of the limited number of samples (only 12), the ability of the calibration model was not fully realized. In this study, 50 samples were created to further test the prediction ability of the calibration model. The conceptual framework is to relate the mass concentrations of the particles in the identified classes to the average aerosol compositions for each sampling time interval using a calibration model based on ART-2a and multivariate analysis. There may be some non-linearity between cluster mass concentrations and ambient species concentrations because of measurement errors, the scaling equations used to estimate particle mass and various assumptions required for building the model. Thus, in this study, PLS regression was integrated with radial basis functions (RBF-PLS) to obtain better prediction effects and compared to partial least square (PLS) regression alone. Compared with an earlier study, these results provide better and a more convincing demonstration of the ability of the calibration model to estimate the chemical compositions from ATOFMS data. The results also suggest that the model would be able to provide carbon data and thus substitute for thermal optical reflectance (TOR) measurements. Additionally, the calibration model based on RBF-PLS showed more accurate predictions in the cases with some non-linearity. Some of the key steps in the modeling effect are also discussed in detail.

show abstract

Section: Description Of Rbf-plsmentioning

confidence: 99%

Predicting bulk ambient aerosol compositions from ATOFMS data with ART-2a and multivariate analysis

Zhao

Hopke

Qin

et al. 2005

Analytica Chimica Acta

View full text Add to dashboard Cite

show abstract

“…In general, RBF neural network is an efficient tool to classify PD signals [30,31]. In this paper, RBF neural network is used as a classifier.…”

Section: Principal Of Rbf Neural Networkmentioning

confidence: 99%

“…On the use of four mathematical models of PD in GIS, the feature space is verified [30]. At last, three feature subsets are inputted into radial base function (RBF) neural network classifier to recognize defects in GIS [31,32]. The results indicate that the feature extraction based on dual-tree CWT is an effective way to extract PD features.…”

Section: Introductionmentioning

confidence: 99%

Feature extraction and recognition of UHF partial discharge signals in GIS based on dual‐tree complex wavelet transform

Tang

Zhou

2009

Int Trans Elec Energy Syst

View full text Add to dashboard Cite

SUMMARYUltra-high-frequency (UHF) partial discharge (PD) detection is an effective means for evaluation of dielectric condition of gas insulated switchgear (GIS). Time-resolved data pattern is utilized to recognize insulation defect because of its attractive advantages that there exists some direct relationship between the shape of PD signal and the physics of the defect. Therefore, the main task is feature extraction of time-resolved data. However, it is hard to extract features in time domain because there is no obvious difference in types of signals. UHF PD signals are non-stationary, so feature extraction in frequency domain is not suitable as well. Feature extraction in time-frequency domain based on discrete wavelet transform (DWT) is a solution for this problem, but there is a weak point which hampers the application of this method, that is, lack of shift invariance, which means that small shifts in the input signal can cause major variations in the distribution of energy between DWT coefficients at different scales. In this paper, a new method of feature extraction of UHF PD signal on the basis of dual-tree complex wavelet transform (CWT) which can reduce shift sensitive is proposed. UHF PD signals are acquired in lab and decomposed by a dual-tree CWT. Then the feature space is constructed through the energy on all frequency bands and modulus maxima on all scales. At last feature space is inputted into a radial base function (RBF) neural network classifier. The results show the superiority of this method over traditional ones.

show abstract

“…In a training process based on the above-mentioned cost function, the patterns that have been correctly recognised still have an unnecessary contribution to the subsequent training process. This is likely to reduce the convergence rate and cause overfitting [10]. In [10], the above issue was addressed and an improved threshold type MSE function was proposed to train RBF networks.…”

Section: Introductionmentioning

confidence: 99%

“…This is likely to reduce the convergence rate and cause overfitting [10]. In [10], the above issue was addressed and an improved threshold type MSE function was proposed to train RBF networks. Together with an appropriate algorithmic scheme, certain improvement on the classification performance was obtained.…”

Section: Introductionmentioning

confidence: 99%

Improving the effectiveness of RBF classifier based on a hybrid cost function

Huang

Chow

2006

Neural Comput & Applic

View full text Add to dashboard Cite

This paper presents a new minimum classification error (MCE)-mean square error (MSE) hybrid cost function to enhance the classification ability and speed up the learning process of radial basis function (RBF)-based classifier. Contributed by the MCE function, the proposed cost function enables the RBF-based classifier to achieve an excellent classification performance compared with the conventional MSE function. In addition, certain learning difficulties experienced by the MCE algorithm can be solved in an efficient and simple way. The presented results show that the proposed method exhibits a substantially higher convergence rate compared with the MCE function.

show abstract

Efficient training of RBF neural networks for pattern recognition

Cited by 43 publications

References 12 publications

Predicting bulk ambient aerosol compositions from ATOFMS data with ART-2a and multivariate analysis

Predicting bulk ambient aerosol compositions from ATOFMS data with ART-2a and multivariate analysis

Feature extraction and recognition of UHF partial discharge signals in GIS based on dual‐tree complex wavelet transform

Improving the effectiveness of RBF classifier based on a hybrid cost function

Contact Info

Product

Resources

About