Showing papers on "Hybrid neural network published in 2005"
TL;DR: Simulations show that the SOFNN has the capability to encode fuzzy rules in the resulting network, based on new adding and pruning techniques and a recursive learning algorithm.
253 citations
TL;DR: In this article, the effects of Pt loading on the performance of the PEM fuel cell have been specifically studied and two hybrid neural network models (multiplicative and additive) have been developed and compared with the full-blown ANN model.
91 citations
TL;DR: A hybrid neural network comprising Fuzzy ARTMAP and FuzzY C-Means Clustering is proposed for pattern classification with incomplete training and test data and the results are analyzed and compared with those from other methods.
Abstract: A hybrid neural network comprising fuzzy ARTMAP and fuzzy c-means clustering is proposed for pattern classification with incomplete training and test data. Two benchmark problems and a real medical pattern classification tasks are employed to evaluate the effectiveness of the hybrid network. The results are analyzed and compared with those from other methods.
59 citations
TL;DR: A two-layer hybrid neural network structure incorporating both supervised and unsupervised learning is proposed to detect and then classify the types of landmines.
Abstract: Neural networks have been applied to landmine detection from data generated by different kinds of sensors. Real-valued neural networks have been used for detecting landmines from scattering parameters measured by ground penetrating radar (GPR) after disregarding phase information. This paper presents results using complex-valued neural networks, capable of phase-sensitive detection followed by classification. A two-layer hybrid neural network structure incorporating both supervised and unsupervised learning is proposed to detect and then classify the types of landmines. Tests are also reported on a benchmark data.
55 citations
TL;DR: In this article, a hybrid neural network model is presented for the simulation of a differential catalytic hydrogenation reactor of carbon dioxide to methanol, which consists of two parts: a mechanistic model and a neural model.
52 citations
TL;DR: In this article, a hybrid neural network model is presented for the simulation of the performance of industrial HDS reactors, which can be used in estimating the catalyst deactivation rate and the impact of feed quality on catalyst activity.
Abstract: A hybrid neural network model is presented for the simulation of the performance of industrial HDS reactors. This model can be used in estimating the catalyst deactivation rate and the impact of feed quality on catalyst activity. A deterministic mathematical code simulating the reactor performance for hydrodesulphurization and hydrogen consumption reactions was used. The deterministic code was coupled with a neural network used to correlate the evaluated kinetic parameters from the industrial data with feed quality and catalyst life time. The neural network is also used to predict the kinetic parameters needed for simulation from the feed quality and the catalyst time on stream. A part of the necessary kinetic parameters were obtained from kinetic experiments performed with the industrial catalyst and with representative feeds in a small scale reactor.
51 citations
18 Mar 2005
TL;DR: One of the first adaptation methods for hybrid systems where the HMM component contributes significantly to the adaptation success is introduced, and a novel approach to the neural network's adaptation is presented, based on the selection of suitable neurons for adaptation.
Abstract: In this paper, strategies are explored to adapt hybrid neural network/HMM systems based on the tied-posterior paradigm. We investigate the retraining of selected important parts of the neural network and a gradient based adaptation strategy for the HMM mixture coefficients based on maximizing the scaled likelihood. The paper presents the following innovations: first, it introduces one of the first adaptation methods for hybrid systems where the HMM component contributes significantly to the adaptation success; second, it presents a novel approach to the neural network's adaptation, based on the selection of suitable neurons for adaptation. Results on the WSJ speaker adaptation test show the capability of our methods to adapt to new speakers, especially in the case of adapting the neural net, and that both methods can be combined to achieve additional improvement of the word error rate in most cases.
38 citations
TL;DR: A novel technique for location prediction of mobile users has been proposed, and a paging technique based on this predicted location is developed to reduce the total location management cost and paging delay, in general.
Abstract: In this paper, a novel technique for location prediction of mobile users has been proposed, and a paging technique based on this predicted location is developed. As a mobile user always travels with a destination in mind, the movements of users, are, in general, preplanned, and are highly dependent on the individual characteristics. Hence, neural networks with its learning and generalization ability may act as a suitable tool to predict the location of a terminal provided it is trained appropriately by the personal mobility profile of individual user. For prediction, the performance of a multi-layer perceptron (MLP) network has been studied first. Next, to recognize the inherent clusters in the input data, and to process it accordingly, a hybrid network composed of a self-organizing feature map (SOFM) network followed by a number of MLP networks has been employed. Simulation studies show that the latter performs better for location management. This approach is free from all unrealistic assumptions about the movement of the users. It is applicable to any arbitrary cell architecture. It attempts to reduce the total location management cost and paging delay, in general.
34 citations
01 Mar 2005
TL;DR: A hybrid learning-based system that integrates neural networks and decision tree learning to overcome the classification problem in a real-time CCP recognition scheme and has better performance in terms of recognition speed and also can accurately identify the type of unnatural CCP.
Abstract: Pattern recognition is an important issue in statistical process control because unnatural patterns displayed by control charts can be associated with specific causes that adversely impact on the manufacturing process. A common problem of existing approaches to control chart pattern (CCP) recognition is false classification between different types of CCP that share similar features in a real-time process-monitoring scenario, in which only limited pattern points are available for recognition. This study proposes a hybrid learning-based system that integrates neural networks and decision tree learning to overcome the classification problem in a real-time CCP recognition scheme. This hybrid system consists of three sequential modules, namely feature extraction, coarse classification, and fine classification. The coarse-classification model employs a four-layer back propagation network to detect and classify unnatural CCPs. The fine-classification module contains four decision trees used in a simple h...
29 citations
24 Jun 2005
26 citations
27 Dec 2005
TL;DR: Simulation results show that this hybrid method outperforms previous approaches, such as mean field annealing, a hybrid of the Hopfield neural network and genetic algorithms, the sequential vertex coloring algorithm, and the gradual neural network.
Abstract: In this paper we propose a hybrid method to solve the broadcast scheduling problem in packet radio networks. In the first stage, we use a backtracking sequential coloring algorithm to obtain a minimal TDMA frame length and the corresponding transmission assignments. In the second stage, we employ the noisy chaotic neural network to find the maximum node transmission based on the results obtained in the previous stage. Simulation results show that this hybrid method outperforms previous approaches, such as mean field annealing, a hybrid of the Hopfield neural network and genetic algorithms, the sequential vertex coloring algorithm, and the gradual neural network.
02 Nov 2005
TL;DR: A hybrid neural network technique is proposed, which consists of the self-organizing map (SOM) and the radial basis function (RBF) network, aiming at optimizing the performance of the recognition and classification of novel attacks for intrusion detection.
Abstract: Intrusion Detection is an essential and critical component of network security systems. The key ideas are to discover useful patterns or features that describe user behavior on a system, and use the set of relevant features to build classifiers that can recognize anomalies and known intrusions, hopefully in real time. In this paper, a hybrid neural network technique is proposed, which consists of the self-organizing map (SOM) and the radial basis function (RBF) network, aiming at optimizing the performance of the recognition and classification of novel attacks for intrusion detection. The optimal network architecture of the RBF network is determined automatically by the improved SOM algorithm. The intrusion feature vectors are extracted from a benchmark dataset (the KDD-99) designed by DARPA. The experimental results demonstrate that the proposed approach performance especially in terms of both efficient and accuracy.
TL;DR: The hybrid model was developed for a river application, using the computational nodes located at the open boundaries to be the ANN nodes for the ANN-FE hybrid model, resulting in savings in computation time.
Abstract: Results obtained from a hybrid neural network-finite element model are reported in this paper. The hybrid model incorporates artificial neural network (ANN) nodes into a numerical scheme, which solves the two-dimensional shallow water equations using finite elements (FE). First, numerical computations are carried out on the entire numerical model, using a larger mesh. The results from this computation are then used to train several preselected ANN nodes. The ANN nodes model the response for a part of the entire numerical model by transferring the system reaction to the location where both models are connected in real time. This allows a smaller mesh to be used in the hybrid ANN-FE model, resulting in savings in computation time. The hybrid model was developed for a river application, using the computational nodes located at the open boundaries to be the ANN nodes for the ANN-FE hybrid model. Real-time coupling between the ANN and FE models was achieved, and a reduction is CPU time of more than 25% was obtained.
26 Oct 2005
TL;DR: This paper presents a new approach for off-line signature verification based on a hybrid neural network (Conic Section Function Neural Network-CSFNN) and compared with two well known neural network architectures.
Abstract: This paper presents a new approach for off-line signature verification based on a hybrid neural network (Conic Section Function Neural Network-CSFNN). Artificial Neural Networks (ANNs) have recently become a very important method for classification and verification problems. In this work, CSFNN was proposed for the signature verification and compared with two well known neural network architectures (Multilayer Perceptron-MLP and Radial Basis Function-RBF Networks). The proposed system was trained and tested on a signature database consisting of a total of 304 signature images taken from 8 different persons. A total of 256 samples (32 samples for each person) for training and 48 fake samples (6 fake samples belonging to each person) for testing were used. The results were presented and the comparisons were also made in terms of FAR (False Acceptance Rate) and FRR (False Rejection Rate).
TL;DR: A high specificity and sensitivity algorithm called PromPredictor for recognizing promoter regions in the human genome that combines a novel promoter recognition model, coding theory, feature selection and dimensionality reduction with machine learning algorithm.
Abstract: This paper proposes a high specificity and sensitivity algorithm called PromPredictor for recognizing promoter regions in the human genome. PromPredictor extracts compositional features and CpG islands information from genomic sequence, feeding these features as input for a hybrid neural network system (HNN) and then applies the HNN for prediction. It combines a novel promoter recognition model, coding theory, feature selection and dimensionality reduction with machine learning algorithm. Evaluation on Human chromosome 22 was ~66% in sensitivity and ~48% in specificity. Comparison with two other systems revealed that our method had superior sensitivity and specificity in predicting promoter regions. PromPredictor is written in MATLAB and requires Matlab to run. PromPredictor is freely available at http://www.whtelecom.com/Prompredictor.htm.
TL;DR: A hybrid neural network that is capable of incremental learning and classification of patterns with incomplete data is proposed, employed as the constituting network for pattern classification while fuzzy c-means clustering is used as the underlying algorithm for processing training as well as test samples with missing features.
Abstract: In this paper, a hybrid neural network that is capable of incremental learning and classification of patterns with incomplete data is proposed. Fuzzy ARTMAP (FAM) is employed as the constituting network for pattern classification while fuzzy c-means (FCM) clustering is used as the underlying algorithm for processing training as well as test samples with missing features. To handle an incomplete training set, FAM is first trained using complete samples only. Missing features of the training samples are estimated and replaced using two FCM-based strategies. Then, network training is conducted using all the complete and estimated samples. To handle an incomplete test set, a non-substitution FCM-based strategy is employed so that a predicted output can be produced rapidly. The performance of the proposed hybrid network is evaluated using a benchmark problem, and its practical applicability is demonstrated using a medical diagnosis task. The results are compared, analysed and quantified statistically with the bootstrap method. Implications of the proposed network for pattern classification tasks with incomplete data are discussed.
07 Nov 2005
TL;DR: The neural network forecasting model based on information entropy and ant colony clustering can be constructed which can effectively reduce the training time and improve convergent speed.
Abstract: This paper presented a hybrid neural network model to integrate information entropy theory and ant colony clustering for load forecasting. As short-term load forecasting is a complex problem with multifactor in power system, if all these factors are used as inputs of neural network, it will not only result in complicated network structure, but also long learning time and inaccurate prediction. First, information entropy theory is used to select relevant ones from all load influential factors, the results are used as inputs of neural network. It can reduce irrelevant load influential factors and the input variables of the input layer for neural network. Next, considering the features of power load and reduced influential factors, using ant colony clustering method, the practical historical load data within one year is divided into several groups. A separate module based on neural networks models each group. Then, the typical samples in each clustered group were selected as the training set for the separate improved Elman neural network which is a kind of globally feed forward locally recurrent network model with distinguished dynamical characteristics. According to the procedures, the reduced input variables and the typical training samples for each neural network can be gotten. Thus the neural network forecasting model based on information entropy and ant colony clustering can be constructed which can effectively reduce the training time and improve convergent speed. During the forecasting process, pattern recognizing is employed to activate the corresponding module for hourly load forecasting. The presented model was tested using Hebei Province daily load data, and the satisfactory results were obtained.
26 Sep 2005
TL;DR: The model of the hybrid neural network is considered, which consists of model ART-2 for clusterization and perceptron for preprocessing of images and provides invariant recognition of objects.
Abstract: The model of the hybrid neural network is considered. This model consists of model ART-2 for clusterization and perceptron for preprocessing of images. The perceptron provides invariant recognition of objects. This model can be used in mobile robots for recognition of new objects or scenes in sight the robot during his movement.
TL;DR: In this paper, a new hybrid-neural-network-based friction component model is developed for powertrain (PT) dynamic analysis and controller design, which has the capability to serve as a forward as well as an inverse system model.
Abstract: In this study, a new hybrid-neural-network-based friction component model is developed for powertrain (PT) dynamic analysis and controller design. This new model, with significantly improved input-output scalability over conventional neural network configuration, has the capability to serve as a forward as well as an inverse system model. The structural information of the available physical and empirical correlations is utilized to construct a parallel-modulated neural network (PMNN) architecture consisting of small parallel sub-networks reflecting specific mechanisms of the friction component engagement process. The PMNN friction component model isolates the contribution of engagement pressure on engagement torque while identifying the nonlinear characteristics of the pressure-torque correlation. Furthermore, it provides a simple torque formula that is scalable with respect to engagement pressure. The network is successfully trained, tested and analyzed, first using analytical data at the component level and then using experimental data measured in a transmission system. The PMNN friction component model, together with a comprehensive powertrain model, is implemented to simulate the shifting process of an automatic transmission (AT) system under various operating conditions. Simulation results demonstrate that the PMNN model can be effectively applied as a part of powertrain system model to accurately predict transmission shift dynamics. A pressure-profiling scheme using a quadratic polynomial pressure-torque relationship of the PMNN model is developed for transmission shift controller design. The results illustrate that the proposed pressure profiling technique can be applied to a wide range of operating conditions. This study demonstrates the potential of the PMNN architecture as a new dynamic system-modeling concept: It not only outperforms the conventional network modeling techniques in accuracy and numerical efficiency, but also provides a new tool for transmission controller design to improve shift quality.
TL;DR: A procedure for pre-processing non-stationary time series is proposed for modelling with a time-delay neural network (TDNN) and uses a fast Fourier transform to determine the TDNN input size.
Abstract: A procedure for pre-processing non-stationary time series is proposed for modelling with a time-delay neural network (TDNN). The procedure stabilises the mean of the series and uses a fast Fourier transform to determine the TDNN input size. Results of applying this procedure on five well-known data sets are compared with existing hybrid neural network techniques, demonstrating improved prediction performance.
22 Jul 2005
TL;DR: A framework for Web page classification is presented that is hybrid architecture of neural network PCA (principle components analysis) and SOFM (self-organizing map) and makes a significant improvement in classifications on both data sets compared with the two conventional methods.
Abstract: Web page classification is one of the essential techniques for Web mining. This paper presents a framework for Web page classification. It is hybrid architecture of neural network PCA (principle components analysis) and SOFM (self-organizing map). In order to perform the classification, a web page is firstly represented by a vector of features with different weights according to the term frequency and the importance of each sentence in the page. As the number of the features is big, PCA is used to select the relevant features. Finally the output of PCA is sent to SOFM for classification. To compare with the proposed framework, two conventional classifiers are used in our experiments: k-NN and Naive Bayes. Our new method makes a significant improvement in classifications on both data sets compared with the two conventional methods.
27 Jun 2005
TL;DR: It is proven that the Preisach-type hysteresis can be transformed to the general continuous mappings such as one- to-one or multi-value-to-one mapping, which can be approximated by the neural network based universal approximators.
Abstract: This paper presents a hybrid neural network (NN) model for hysteresis in mechanical or piezoelectric systems. It is proven that the Preisach-type hysteresis can be transformed to the general continuous mappings such as one-to-one or multi-value-to-one mapping, which can be approximated by the neural network based universal approximators. The proposed hybrid neural model consists of two neural networks, i.e. a double-threshold neural network (DTNN) is proposed to memorize the historic information of the input; after that a multi-layer neural network (MNN) is utilized to approximate hysteresis nonlinearity based on the information stored in the DTNN
TL;DR: This paper proposes a comparison between several hybrid models based on the two most widespread neural networks, the MultiLayer Perceptron and the Radial Basis Function network, based on simulations of fed-batch bacterial cultures.
25 Jul 2005
TL;DR: A new genetic granular cognitive fuzzy neural network based on granular computing, soft computing and cognitive science is used in a pattern recognition problem to compare human brains with the biological neural networks.
Abstract: Biological neural networks in the human brain can recognize different patterns with noise by the unknown biologically cognitive pattern recognition method. Since the human brain consists of biological neural networks that are the major components performing pattern recognition, it is very interesting and very important to investigate how the biological neural networks and the artificial neural networks recognize different patterns. A new genetic granular cognitive fuzzy neural network based on granular computing, soft computing and cognitive science is used in a pattern recognition problem to compare human brains with the biological neural networks. The hybrid genetic forward-wave-backward-wave learning algorithm is used to enhance learning quality. Both pattern recognition results generated by human persons and the genetic granular cognitive fuzzy neural network are analyzed in terms of computer science and cognitive science.
Journal Article•
TL;DR: In this article, a nonlinear self-adaptive predictive function control (PFC) based on a hybrid neural network is presented, which is composed of BP network and linear neural network.
Abstract: A basic predictive function control(PFC) is only applicable for linear plant control.To overcome the defect,a nonlinear self-adaptive PFC based on a hybrid neural network is presented.The hybrid neural network is composed of BP network and linear neural network.One can identify a nonlinear plant described with Hammerstein model effectively by the hybrid neural network.The BP network reflects the nonlinear static gain .The linear neural network reflects the linear dynamic subsystem.Then the inverse form of nonlinear static gain is solved and in series with the nonlinear plant to compensate the nonlinear static gain of nonlinear plant.Thus the nonlinear plant control is transformed into linear plant control and the PFC of nonlinear plant is realized.The control system can adjust the weights of hybrid neural network and the parameters of controller timely to keep good control performance when the character of controlled plant varies.Simulation results show that the control system has good control effect.
TL;DR: A novel fuzzy neural network model based on fuzzy clustering method that can accept continuous and discrete inputs together and is superior to the traditional neuralnetwork model in vision-based sensors is proposed.
Abstract: This paper proposes a novel fuzzy neural network model based on fuzzy clustering method. The model can accept continuous and discrete inputs together; the discrete input to the model is divided into several clusters by using fuzzy c-mean clustering algorithm (FCM). A fuzzy clustering neuron (FC-neuron) is designed to calculate a membership degree value belonging to one cluster for each discrete input. A four-layer hybrid neural network is constructed; fuzzy-neurons and FC-neurons construct the antecedent part of fuzzy rules. A multi-input multioutput hybrid neural network was designed by the novel modeling method and applied to vision-based sensors. Simulation results show this method is superior to the traditional neural network model in vision-based sensors.
TL;DR: This paper first presents a discussion of the need to incorporate “intelligence” into an automated design process and the various constraints that designers face when embarking on industrial design projects, and presents the design problem as optimizing the design output against constraints and the use of soft computing and hybrid intelligent systems techniques.
Abstract: Contemporary design process requires the development of a new computational intelligence or soft computing methodology that involves intelligence integration and hybrid intelligent systems for design, analysis and evaluation, and optimization. This paper first presents a discussion of the need to incorporate “intelligence” into an automated design process and the various constraints that designers face when embarking on industrial design projects. Then, it presents the design problem as optimizing the design output against constraints and the use of soft computing and hybrid intelligent systems techniques. In this paper, a soft-computing-integrated intelligent design framework is developed. A hybrid dual cross-mapping neural network (HDCMNN) model is proposed using the hybrid soft computing technique based on “cross-mapping” between a back-propagation network (BPNN) and a recurrent Hopfield network (HNN) for supporting modeling, analysis and evaluation, and optimization tasks in the design process. The two networks perform different but complementary tasks—the BPNN “decides” if the design problem is a “type 0” (rational) or “type 1” (non-rational) problem, and the output layer weights are then used as the energy function for the HNN. The BPNN is used for representing design patterns, training classification boundaries, and outputting network weight values to the HNN, and then the HNN uses the calculated network weight values to evaluate and modify or re-design the design patterns. The developed system provides a unified soft-computing-integrated intelligent design framework with both symbolic and computational intelligence. The system has self-modifying and self-learning functions. Within the system, only one network training is needed for accomplishing the evaluation, rectification/modification, and optimization tasks in the design process. Finally, two case studies are provided to illustrate and validate the developed model and system.
27 Dec 2005
TL;DR: The stability and the bifurcation of the neuron cluster and factors affecting cluster generation are investigated.
Abstract: In this paper, dynamics of VSF-network (vibration synchronizing function network) is investigated. VSF-network is a model of neural networks that segments information from external world and fixes the segmented information. VSF-network is a hybrid neural network, and the chaos neuron is used for the hidden layer of it. VSF-network articulates information with the neuron cluster generated by the synchronizing vibration that the chaos neuron in hidden layer shows. We analyze the dynamics of generating the neuron cluster. Factors affecting cluster generation are investigated. The stability and the bifurcation of the neuron cluster and factors affecting cluster generation are investigated.
10 Oct 2005
TL;DR: The simulation results show that the proposed cooperative control approach that is based on the combination of neural network and cerebellar model articulation controller methodology not only drastically reduces the overshoot, but also maintains a small extent of the settling time and the steady state error.
Abstract: In this paper, we propose a cooperative control approach that is based on the combination of neural network (NN) and cerebellar model articulation controller (CMAC) methodology. The main purpose is not only the chattering free but also the two inverted pendulums on carts a superior tracking response. Moreover, the system performance obtained via the method of hybrid sliding mode control can be improved. In the present approach, two parallel networks are utilized to realize a hybrid neural sliding mode control (HNSMC). The equivalent control and the corrective control in terms of sliding mode control are the outputs of the neural networks. The weights adaptations of neural network are determined based on the sliding mode control equations. The simulation results show that the proposed controller not only drastically reduces the overshoot, but also maintains a small extent of the settling time and the steady state error. It is shown that the proposed method is feasible and effective.
28 Nov 2005
TL;DR: This work uses the self-organized neural network to deal with clustering analysis and character process of sinter process, and the hybrid neural network is used to quickly predict the BTP of the subclass to improve the response of system and satisfy the real-time control requirement.
Abstract: Due to time varying, long time delay, noise and multimode, we use the self-organized neural network to deal with clustering analysis and character process of sinter process, and use the hybrid neural network to quickly predict the BTP of the subclass to improve the response of system and satisfy the real-time control requirement. Through experimental testing and data analyzing, the system possess stronger robust and adaptable