Showing papers on "Hybrid neural network published in 1996"

Hybrid neural network models for bankruptcy predictions

Abstract: The objective of this paper is to develop the hybrid neural network models for bankruptcy prediction. The proposed hybrid neural network models are (1) a MDA-assisted neural network, (2) an ID3-assisted neural network, and (3) a SOFM(self organizing feature map)-assisted neural network. Both the MDA-assisted neural network and the 11)3-assisted neural network are the neural network models operating with the input variables selected by the MDA method and 1133 respectively. The SOFM-assisted neural network combines a backpropagation model (supervised learning) with a SOFM model (unsupervised learning). The performance of the hybrid neural network model is evaluated using MDA and ID3 as a benchmark. Empirical results using Korean bankruptcy data show that hybrid neural network models are very promising neural network models for bankruptcy prediction in terms of predictive accuracy and adaptability.

Method and apparatus for object recognition

Abstract: A hybrid neural network system for object recognition exhibiting local image sampling, a self-organizing map neural network, and a hybrid convolutional neural network. The self-organizing map provides a quantization of the image samples into a topological space where inputs that are nearby in the original space are also nearby in the output space, thereby providing dimensionality reduction and invariance to minor changes in the image sample, and the hybrid convolutional neural network provides for partial invariance to translation, rotation, scale, and deformation. The hybrid convolutional network extracts successively larger features in a hierarchical set of layers. Alternative embodiments using the Karhunen-Loeve transform in place of the self-organizing map, and a multi-layer perceptron in place of the convolutional network are described.

Fuzzy neural network and fuzzy expert system for load forecasting

Abstract: A hybrid neural network fuzzy expert system is developed to forecast short-term electric load accurately. The fuzzy membership values of the load and other weather variables are the inputs to the neural network, and the output comprises the membership values of the predicted load. An adaptive fuzzy correction scheme is used to forecast the final load by using a fuzzy rule base and fuzzy inference mechanism. Extensive studies have been performed for all seasons, and a few examples are presented in the paper, including average, peak and hourly load forecasts.

Convolutional neural networks for face recognition

Abstract: Faces represent complex, multidimensional, meaningful visual stimuli and developing a computational model for face recognition is difficult. We present a hybrid neural network solution which compares favorably with other methods. The system combines local image sampling, a self-organizing map neural network, and a convolutional neural network. The self-organizing map provides a quantization of the image samples into a topological space where inputs that are nearby in the original space are also nearby in the output space, thereby providing dimensionality reduction and invariance to minor changes in the image sample, and the convolutional neural network provides for partial invariance to translation, rotation, scale, and deformation. The method is capable of rapid classification, requires only fast, approximate normalization and preprocessing, and consistently exhibits better classification performance than the eigenfaces approach on the database considered as the number of images per person in the training database is varied from 1 to 5. With 5 images per person the proposed method and eigenfaces result in 3.8% and 10.5% error respectively. The recognizer provides a measure of confidence in its output and classification error approaches zero when rejecting as few as 10% of the examples. We use a database of 400 images of 40 individuals which contains quite a high degree of variability in expression, pose, and facial details.

Adaptive propagation modelling using a hybrid neural network technique

Abstract: A hybrid modelling approach for the prediction of terrestrial wave propagation is introduced. It is shown that the use of neural networks leads to highly adaptive models by the simultaneous representation of nonlinear dependencies of various environmental parameters. This flexible and computationally effective approach can be used for calibration and as an extension of conventional prediction models.

Spectroscopy and hybrid neural network analysis

Abstract: This paper reviews the current use of spectroscopy and related instrumentation in chemical analysis. Advancements in digital signal processing technology are making it possible to improve the sensitivity and accuracy of analytical instruments without expensive upgrading of instrument hardware. A hybrid neural network (HNN) is described that can perform nonlinear signal analysis. The HNN approach combines the simple data reduction capability of conventional linear signal processing algorithms with the adaptive learning and recognition ability of a multilayer nonlinear neural network architecture. A number of examples show the rise of the HNN for environmental monitoring and real-time process control.

A hybrid neural network-fuzzy logic limit protection system for rotorcraft

Abstract: This paper describes the development of a hybrid neural network/fuzzy logic system for detection and prevention of helicopter flight limit exceedances. The hybrid system uses neural networks to model several limits associated with the UH-1H helicopter. Fuzzy logic (FL) algorithms characterize the aircraft's flight condition with respect to these limits, and provide a continuous measure of limit exceedance. This approach offers more flexibility than a binary classification of limit exceedance (e.g., 0 for not exceeded, 1 for exceeded), since a continuous measure of cxcccdance with a specified threshold can warn of impending exceedances, particularly through the use of state rate information. This approach to characterizing limit exceedance can support a wide variety of pilot cueing modalities, including direct control intervention, In this study, the FL-derived measures drive control blending logic that mix pilot inputs with limit protection control inputs designed to prevent exceedances. The hybrid approach was implemented and tested in an off-line engineering six degree-of-freedom simulation model of the UH-1H. Results show that the approach is feasible and that it offers considerable potential for implementing onboard limit protection in rotorcraft.

Impact location, energy, and damage detection in a smart structure

Abstract: A sensor at a fixed location in a complex structure records a complicated but unique wave pattern containing information about impact location, imparted energy and any damage created by an impact event. An intelligently designed hybrid neural network system is capable of extracting this information from the sensory signal. Such a system based on a generalized regression neural network (GRNN) is described for the purpose of impact location, energy, and damage detection. A Northrop Grumman test article is utilized to demonstrate capabilities of the system. The system performance evaluation based on the preliminary experiments is very encouraging. Further experimental evaluations of the system are planned and are described in this manuscript.© (1996) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Design and implementation of a parallel software for hybrid neural network computation in PVM environment

Abstract: This paper discusses the design of a hybrid artificial neural network (ANN) system and its implementation in the Parallel Virtual Machine (PVM) environment. First, the PVM functions for supporting parallel applications and communications among multiple processes and multiple machines are investigated. Then, the design and construction of a hybrid ANN simulation software is proposed. It includes the user interface, control and SPMD computing levels. The software can be used for supporting parallel simulation of different kinds of learning algorithms and neural computing models.

Hybrid BRAINNE: a method for developing symbolic disjunctive rules from a hybrid neural network

Abstract: A method for learning disjunctive rules using a combination of two existing neural network schemes is proposed. The hybrid network consists of two layers; the first is an unsupervised network while the second is a supervised network. The first layer is used for ordering the inputs of training instances into clusters. Initial rules are extracted from this layer using an existing technique called Unsupervised BRAINNE. These rules are then fed into the second layer which is trained using the delta rule. The second layer is then examined to determine which clusters define the output nodes. This method is able to identify disjunctive rules directly rather than utilising a generate and test paradigm as was used in previous supervised versions of BRAINNE.

Hybrid neural networks and network design

Abstract: The major contribution of this dissertation is to design and demonstrate the effectiveness of neural networks hybridized with problem specific meta-knowledge to solve combinatorial optimization problems. Heuristic procedures are incorporated into the neural network topological design. The learning rules implemented within the neural network adjust the cost weights to improve on the initial heuristic results obtained by the neural network. The viability of the neural network implementation is shown by solving the Capacitated Minimum Spanning Tree (CMST) and the Tree-Star telecommunications network topology design problems. This dissertation also modifies the Hopfield neural network topology to solve the Traveling Salesperson Problem and the CMST problem. Computational results for the Hopfield implementation of the CMST, both in terms of computational time and quality of solution, are not as good as the results for the hybrid neural network techniques originated in this dissertation.

Performance analysis of a wavelet-based hybrid neurosystem for signal classification

Abstract: This paper is concerned with the problem of determining performance of a wavelet-based hybrid neurosystem trained to provide efficient feature extraction and signal classification. The hybrid networkconsists of a parallel array of neurosystems. Each neurosystem is constructed with three single neuralnetworks; two ofwhich arefeature extraction networks, and the other is a classification network The twof eature extraction networks, namely, the magnitude network and location network, are provided withmagnitude and location information of the wavelet transform coefficients, respectively, and are trainedwith self-organizing rules. Their outputs are then presented to the classification network for patternrecognition. Based on the topological maps provided by thefeature extraction neural networks, the back-propagation algorithm is used to train the third network for pattern recognition. The combination ofwavelet, wavelet transform, and hybrid neural network architecture and advanced training algorithms inthe design makes the system unique and provides high classification accuracy. In this paper, systempeiformance is shown to be intrinsically related to basis kernel function used in feature extraction. Amethod for selecting the optimal basis function and a performance analysis using simulated data undervarious noise condition are presented and compared against other pattern recognition techniques.Keywords: wavelets, hybrid neurosystem, hybrid architecture, signal classifications, pattern recognition1. INTRODUCTIONMethods of pattern recognition can be applied to a wide range of applications, including signalcharacterization, feature extraction, image compression, and signal classification. In particular, applicationsin signal classification require efficient and robust pattern recognition techniques. Signal classificationinvolves the extraction and partitioning of features of targets of interest. In many situations, the problem iscomplicated by the uncertainty of the signal origin, fluctuations in the presence of noise, the degree ofcorrelation of multi-sensor data, and the interference of nonlinearities in the environment. In the last severaldecades, methods for pattern recognition have been developed from traditional signal processin techniquessuch as correlation and frequency characterization,' to artificial neural network technology. Recently,research in signal processing has achieved new improvements with advanced pattern recognition techniques,and more importantly, has found a new way of transforming signals -

Application of a hybrid neural network for the discrimination of seeds by artificial vision

Abstract: Intelligent hybrid systems are playing an increasing role in the development of artificial intelligence. In this study, we applied simulated annealing to adjust the weights of a multilayer neural network (MNN). Two versions of simulated annealing were tested: conventional simulated annealing (CSA) and fast simulated annealing (FSA). The applied hybrid system was used as a classifier in order to discriminate between 3 seed species (1 cultivated seed species which is perennial rye grass, and 2 adventitious seed species which are rumex and wild oat). From a set of colour digital images, 73 morphometrical and textural features were extracted to characterise each individual seed. Stepwise discriminant analysis made it possible to select the first 3 relevant features. The performances of classification were highly dependent on the scaling parameters of simulated annealing. For example, when the number of iterations of simulated annealing was 5, and the number of temperatures was 40, the combination between CSA and MNN correctly classified 98.18 and 97.77 percent of the training and the test sets, whereas FSA and MNN identified 99.18 and 99.68 percent of the same data sets. Globally, FSA outperformed CSA both in reliability and computational resources. A hybrid system combined with a colour image analysis showed promise for the design of an automatic seed identification device.

Adaptive-sized hybrid neural network for segmentation of breast cancer cells in pathology images

Abstract: Akira Hasegawa', Kevin J. Cullen2, Seong K. Mu&1 Georgetown University Medical Center, Department of Radiology2115 Wisconsin Ave., NW, Suite 603, Washington, DC 200072 Georgetown University Medical Center, Division of Medical OncologyABSTRACTIn this report, we describe a novel method to automatically segment several kinds of cells in breast cancerpathologyimages. The information on the number of cells is expected to assist pathologists in consistent diagnosis of breast cancer.Currently, most pathologists make a diagnosis based on a rough estimation of the number of cells on an image. Because of therough estimation, the diagnosis is not objective. To assist pathologists to make a consistent, objective and fast diagnosis, it isnecessary to develop a computer system to automatically recognize and count several kinds of cells. As first step of this purpose,we proposed a novel neural network model, called an adaptive-sized hybrid neural network (ASH-NN), and developed a method,based on this network model, to segment cells from breast cancer pathology images. The proposed neural network consists ofthree layers and the connection weights between the first and second layers are updated by self-organization, and the weightsbetween the second and third layers are determined based on supervised learning. The ASH-NN has the capability of (1) automaticadjustment of the number of hidden units and (2) quick learning.Keywords: neural networks, self-organization, color image segmentation, breast cancer, pathology image1. INTRODUCTIONDiagnosis of breast cancer based on pathology images is commonly done to identify malignancy of suspicious tumors ormicrocalcifications. Fig. 3 (a) shows an example of a breast cancer pathology image. Tissue sections of breast cancer are stained,and magnified by a microscope. In Fig. 3 (a), the rounded cell indicated by the red arrow is an epithelial cell, which is a cancercell, a long cell indicated by the blue arrow is a stromal cell, and the small black dots overall the image are insulin-like growthfactor-Il (IGF-II) mRNAs. Generally, IGF-II mRNA is a potent mitogen for a variety of cell types and is considered an importantregulator of breast cancer growth [1]. This means that cells overlapped with clustered IGF-II mRNAs are active and growing. Inthe treatment of breast cancer, it is important to recognize and count active cancer cells. However, without the help ofcomputers,counting cells is time-consuming work. Currently, most pathologists make a diagnosis based on a rough estimation of thenumber of cells on an image. Because of the rough estimation, the diagnosis is not objective. To assist pathologists to make aconsistent diagnosis, it is necessary to develop a computer system to automatically recognize and count several kinds of cells.The system will provide pathologists with not only a consistent but also fast diagnosis.A study on automatic segmentation of breast tissues by a back propagation neural network [2] has been reported by Okiiet al. [3]. However, they used input images with a notable difference only in brightness, but not color. In addition, the numberof the training examples used for training of a back propagation neural network was so small that the trained neural network'sperformance was not expected to be generalized. Generally, it is said that to get a generalized performance of a back propagationneural network, the necessary number of training examples is more than ten times of the number of connection weights in theneural network [4, 5]; nevertheless, Okii et al. used only 48 training examples for training a neural network with more than 360connection weights.For the image segmentation of general color images, several techniques have been reported. Methods based on thethresholding of image histograms have been proposed [6-8]. In these methods, on each color axis, color space is divided by meansof histogram thresholding, similar to a technique used for the segmentation on gray-scale images. Other techniques based on theK-means algorithm and the least sum of squares criterion have been reported [9, 10]. Recently, a method using neural network hasalso been proposed [1 1]. The basic idea of this method is a vector quantization, which is done by a competitive learning. Imagesegmentation is done by clustering color space by the vector quantization. In these methods, however, the number of segments isusually not specified and depends on the pathology image. On the other hand, in segmentation on pathology images, the numberof segments is assigned.

CMHNN: a constructive modular hybrid neural network for classification

Abstract: We propose a constructive RBF-like network that is able to learn discriminant functions in a multiclass classification problem where patterns are not individually labeled, but they belong to a higher level structure where knowledge about classes is present. The main differences with the standard RBF approaches can be summarized in two points. The number of localized receptive field (LRF) units is not fixed beforehand. Instead of it, we create a modular hidden layer with a constructive criteria that allows adding and updating units to each module. The supervised learning procedure doesn't search for a minimum of the error function; it is a decision-based method that updates the connections from each hidden module to the output and affects the creation of LRF units. This architecture has rendered very good results on the classification of real images drawn from the database created for the ALINSPEC project.

Comparison of linear and morphological shared-weight neural networks

Abstract: A hybrid neural network that can learn nonlinear morphological feature extraction and classification simultaneously, called the morphological shared-weight network (MSNN), is described. The feature extraction operation is performed by a gray scale hit-miss transform. The network learns morphological structuring elements by a back-propagation type learning rule. It provides a general problem-independent methodology for designing morphological structuring elements for pattern recognition. The network was applied to handwritten digit recognition and automatic target recognition (ATR) of occluded vehicles and compared to the standard shared-weight neural networks (SSNN) that perform linear feature extraction. For binary handwritten digit recognition, it produced performance comparable to that obtained using existing shared-weight networks. However, it trained faster. For ATR, a set of parking lot images containing a certain type of vehicle was used. An MSNN was trained with non- occluded training vehicles and tested with images containing the training vehicles at various degrees of occlusion. An efficient training method to improve background rejection is introduced. Two target-aim-point selection methods are defined. The MSNN performed significantly better than the SSNN at detecting occluded vehicles and reducing false alarm rates. Furthermore, the morphological network trained significantly faster.

k and fuzz

Abstract: Indexing terms: Neural networks, Load forecasting, Expert system, Power system planning Abstract: A hybrid neural network fuzzy expert system is developed to forecast short-term electric load accurately. The fuzzy membership values of the load and other weather variables are the inputs to the neural network, and the output comprises the membership values of the predicted load. An adaptive fuzzy correction scheme is used to forecast the final load by using a fuzzy rule base and fuzzy inference mechanism. Extensive studies have been performed for all seasons, and a few examples are presented in the paper, including average, peak and hourly load forecasts.

Hydraulic Servocylinder Position Control Using a Hybrid Neurofuzzy Controller

Abstract: The purpose of this article is to design a neurofuzzy controller with a hybrid learning algorithm and to control the position of a hydraulic servocylinder with an IBM compatible microcomputer. The structure of the neurofuzzy controller is based on the bell-shaped membership function and the Mamdani fuzzy reasoning rules. According to the training data and the hybrid neural network learning, the minimum fuzzy reasoning rules and the optimized membership function can be found automatically. The effects of different design parameters and the load disturbance are also studied experimentally.