Showing papers on "Hybrid neural network published in 2010"

Conceptual cost estimates using evolutionary fuzzy hybrid neural network for projects in construction industry

Abstract: Conceptual cost estimates are important to project feasibility studies and impact upon final project success. Such estimates provide significant information that can be used in project evaluations, engineering designs, cost budgeting and cost management. This study proposes an artificial intelligence approach, the evolutionary fuzzy hybrid neural network (EFHNN), to improve conceptual cost estimate precision. This approach first integrates neural networks (NN) and high order neural networks (HONN) into a hybrid neural network (HNN), which operates with alternating linear and non-linear neuron layer connectors. Fuzzy logic (FL) is then used in the HNN to handle uncertainties, an approach that evolves the HNN into a fuzzy hybrid neural network (FHNN). As a genetic algorithm is employed on the FL and HNN to optimize the FHNN, the final version used for this study may be most aptly termed an 'EFHNN'. For this study, estimates of overall and category costs for actual projects were calculated and compared. Results showed that the proposed EFHNN may be deployed effectively as an accurate cost estimator during the early stages of construction projects. Moreover, the performance of linear and non-linear neuron layer connectors in EFHNN surpasses models that deploy a singular linear NN.

Population Balance Model-Based Hybrid Neural Network for a Pharmaceutical Milling Process

Abstract: Population balances are generally used to predict the particle size distribution resulting from the processing of a particulate material in a milling unit. The key component of such a model is the breakage function. In this work we present an approach to model breakage functions that has utility for situations in which determination of the breakage function from first principles is difficult. Traditionally, heuristic models have been used in those situations but the unstructured nature of such models limits their applicability and reliability. To address this gap, we propose a semi-empirical hybrid model that integrates first principles knowledge with a data-driven methodology that takes into account the material properties, mill characteristics, and operating conditions. The hybrid model combines a discrete form of population balance model with a neural network model that predicts the milled particle size distribution given material and mill information. We demonstrate the usefulness of this approach for compacted API ribbons milled in a lab scale Quadra conical mill for different materials and mill conditions. Comparisons are also given to the predictions obtained via a purely neural network model and a population balance model with a linear breakage kernel.

A hybrid neural network–genetic algorithm approach for permutation flow shop scheduling

Abstract: The objective of this paper is to find a sequence of jobs for the permutation flow shop to minimise the makespan. The shop consists of 10 machines. A feed-forward back-propagation artificial neural network (ANN) is used to solve the problem. The network is trained with the optimal sequences for five-, six- and seven-job problems. This trained network is then used to solve a problem with a greater number of jobs. The sequence obtained using the neural network is used to generate the initial population for the genetic algorithm (GA) using the random insertion perturbation scheme (RIPS). The makespan of the sequence obtained by this approach (ANN-GA-RIPS) is compared with that obtained using GA starting with a random population (ANN-GA). It was found that the ANN-GA-RIPS approach performs better than ANN-GA starting with a random population. The results obtained are compared with those obtained using the Nawaz, Enscore and Ham (NEH) heuristic and upper bounds of Taillard's benchmark problems. ANN-GA-RIPS per...

Evolutionary Approach for Mobile Robot Path Planning in Complex environment

Abstract: The shortest/optimal path planning in a static environment is essential for the efficient operation of a mobile robot. Recent advances in robotics and machine intelligence have led to the application of modern optimization method such as the genetic algorithm (GA), to solve the path-planning problem. In this paper, the problem of finding the optimal collision free path in complex environments for a mobile robot is solved using a hybrid neural network, Genetic Algorithm and local Search method. We constructed the neural network model of environmental and used this model to establish the relationship between a collision avoidance path and the output of the model. What is new in this work is a novel representation of solutions for evolutionary algorithms that is efficient, simple and also compatible with Hybrid algorithm. The new representation makes it possible to solve the problem with a small population and in a few generations. It also makes the genetic operator simple and allows using an efficient local search operator within the evolutionary algorithm. The performance of the proposed GA approach is tested on eight different environments consisting of polygonal obstacles with increasing complexity.

A hybrid neural network strategy for identification of structural parameters

Abstract: This paper presents a multistage identification scheme for structural damage detection using modal data. Previous studies of damage assessment using neural networks mostly involved training a backpropagation neural network (BPN) to learn damage patterns that were obtained either experimentally or by simulation for different damage cases. Damage identification for large structures, especially those involving multiple member damage, could result in large training data sets that require a large BPN and consequently greater computational effort. The proposed scheme involves using a counterpropagation neural network (CPN) in the first stage for sorting the training data into clusters and giving an approximate guess of the damage extent within a very short time. After an approximate estimate of the damage is obtained, a new set of training patterns of reduced size is generated using the CPN prediction. In the second stage, a BPN trained with the Levenberg–Marquardt algorithm is used to learn the new training da...

A hybrid neural network model for cyanobacteria bloom in Dianchi Lake

Abstract: Cyanobacteria bloom predicting is an important part of water quality management in eutrophic lakes or reservoirs This paper developed a hybrid model consisting of back-propagation neural network and rough decision to predict the cyanobacteria bloom in Dianchi Lake using weather conditions The rough reduct could be used to select essential factors for the neural network The training efficacy of the hybrid model was more effective than that of neural network model merely And compared to other models, the predicting accuracy of the hybrid model was also obviously improved

Hybrid modeling of inulinase bio-production process

Abstract: BACKGROUND: A potential application of inulinase in the food industry is the production of fructooligosaccharides (FOS) through transfructosilation of sucrose. Besides their ability to increase the shelf-life and flavor of many products, FOS have many interesting functional properties. The use of an industrial medium may represent a good, cost-effective alternative to produce inulinase, since the activity of the enzyme produced may be improved or at least remain the same compared with that obtained using a synthetic medium. Thus, inulinase production for use in FOS synthesis is of considerable scientific and technological appeal, as is the development of a reliable mathematical model of the process. This paper describes a hybrid neural network approach to model inulinase production in a batch bioreactor using agroindustrial residues as substrate. The hybrid modeling makes use of a series artificial neural network to estimate the kinetic parameters of the process and the mass balance as constitutive equations. RESULTS: The proposed model was shown to be capable of describing the complex behavior of inulinase production employing agroindustrial residues as substrate, so that the mathematical framework developed is a useful tool for simulation of this process. CONCLUSION: The hybrid neural network model developed was shown to be an interesting alternative to estimate model parameters since complete elucidation of the phenomena and mechanisms involved in the fermentation is not required owing to the black-box nature of the ANN used as parameter estimator. Copyright © 2010 Society of Chemical Industry

Hybrid-maximum neural network for depth analysis from stereo-image

Abstract: In present paper, we describe completely innovation architecture of artificial neural nets based on Hopfield structure for solving of stereo matching problem. Hybrid neural network consists of classical analogue Hopfield neural network and maximal neural network. The role of analogue Hopfield network is to find of attraction area of global minimum, whereas maximum network is to find accurate location of this minimum. Presented network characterizes by extremely high rate of working with the same accuracy as classical Hopfield-like network. It is very important as far as application and system of visually impaired people supporting are concerned. Considered network was taken under experimental tests with using real stereo pictures as well simulated stereo images. This allows on calculation of errors and direct comparison to classic analogue Hopfield neural network. Results of tests have shown, that the same accuracy of solution as for continuous Hopfield-like network, can be reached by described here structure in half number of classical Hopfield net iteration.

Hybrid Neural Network Intrusion Detection System Using Genetic Algorithm

Abstract: In this paper, we introduce an Intrusion Detection system (IDS) based Hybrid Evolutionary Neural Network (HENN). A brief overview of IDS, genetic algorithm, and related detection techniques are discussed. The system architecture is also introduced. Factors affecting the genetic algorithm are addressed in detail. Unlike other implementations of IDS, Input features, network structure and connection weights are evolved using genetic algorithm in HENN. This is helpful for identification of complex anomalous behaviors. Experimental results show that the proposed IDS can efficiently improve the detection rate and correctness rate.

Speech recognition of deaf and hard of hearing people using hybrid neural network

Abstract: This paper describes isolated word recognition of deaf students by unsupervised and supervised neural network. Compared to normal speech, there is high variability in deaf speech and by hearing once we couldn't understand it. By the use of proposed method deaf people can make use of all voice operated devices. In this paper we use combination of SOFM and BPN neural network for recognition. Initially the input is sampled, filtered, windowed and Perceptual Linear Predictive Coefficients are determined for each frame. These coefficients are applied as input to the SOFM neural network. The output of this network is given to BPN neural network comprising of 3 layers for learning. The network has been trained with five words uttered by five different deaf persons in the age group of 5–10 years. Another set of same five words uttered by same five deaf persons were used for test purposes. The recognition results for the word one, three, four is 50 to 60% and for five is 50%…But the recognition results for word two are only10% since the variability is high for two. The results can be improved by varying the parameters of the hybrid neural network.

Hybrid neural networks as prediction models

Abstract: The paper presents hybrid neural networks as prediction models for water intake in water supply system. Previous research concerned establishing prediction models in the form of single neural networks: linear network (L), multi-layer network with error back propagation (MLP) and Radial Basis Function network (RBF). Currently, the models in the form of hybrid neural networks (L-MLP, L-RBF, MLP-RBF and L-MLP-RBF) were created. The prediction models were compared for obtaining optimal prognosis. Prediction models were done for working days, Saturdays and Sundays. The research was done for selected nodes of water supply system: detached house node and nodes for 4 hydrophore stations from different pressure areas of water supply system. Models for Sundays were presented in detail.

Large-Scale Pattern Storage and Retrieval Using Generalized Brain-State-in-a-Box Neural Networks

Abstract: In this paper, a generalized Brain-State-in-a-Box (gBSB)-based hybrid neural network is proposed for storing and retrieving pattern sequences The hybrid network consists of autoassociative and heteroassociative parts Then, a large-scale image storage and retrieval neural system is constructed using the gBSB-based hybrid neural network and the pattern decomposition concept The notion of the deadbeat stability is employed to describe the stability property of the vertices of the hypercube to which the trajectories of the gBSB neural system are constrained Extensive simulations of large scale pattern and image storing and retrieval are presented to illustrate the results obtained

Content-based hierarchical document organization using multi-layer hybrid network and tree-structured features

Abstract: Automatic organizing documents through a hierarchical tree is demanding in many real applications. In this work, we focus on the problem of content-based document organization through a hierarchical tree which can be viewed as a classification problem. We proposed a new document representation to enhance the classification accuracy. We developed a new hybrid neural network model to handle the new document representation. In our document representation, a document is represented by a tree-structure that has a superior capability of encoding document characteristics. Compared to traditional feature representation that encodes only global characteristics of a document, the proposed approach can encode both global and local characteristics of a document through a hierarchical tree. Unlike traditional representation, the tree representation reflects the spatial organizations of words through pages and paragraphs of a document that help to encode better semantics of a document. Processing hierarchical tree is another challenging task in terms of computational complexity. We developed a hybrid neural network model, composed of SOM and MLP, for this task. Experimental results corroborate that our approach is efficient and effective in registering documents into organized tree compared with other approach.

Hybrid neural network for prediction of CO2 solubility in monoethanolamine and diethanolamine solutions

Abstract: The solubility of CO2 in single monoethanolamine (MEA) and diethanolamine (DEA) solutions was predicted by a model developed based on the Kent-Eisenberg model in combination with a neural network. The combination forms a hybrid neural network (HNN) model. Activation functions used in this work were purelin, logsig and tansig. After training, testing and validation utilizing different numbers of hidden nodes, it was found that a neural network with a 3-15-1 configuration provided the best model to predict the deviation value of the loading input. The accuracy of data predicted by the HNN model was determined over a wide range of temperatures (0 to 120 °C), equilibrium CO2 partial pressures (0.01 to 6,895 kPa) and solution concentrations (0.5 to 5.0M). The HNN model could be used to accurately predict CO2 solubility in alkanolamine solutions since the predicted CO2 loading values from the model were in good agreement with experimental data.

A Hybrid Neural Network and Genetic Algorithm Model for Predicting Dissolved Oxygen in an Aquaculture Pond

Abstract: The prediction for dissolved oxygen (DO) in aquaculture ponds is a problem of multi-variables, nonlinearity and long-time lag. Neural networks (NNs) have become one of ideal tools in modeling nonlinear relationship between inputs and outputs. In this work, GA-LM, a neural network model combining Levenberg–Marquardt(LM) algorithm and Genetic Algorithm (GA) was developed for predicting DO in an aquaculture pond at Dalian, China. LM was used to train NNs, showing faster convergence rate. The network architecture was optimized by GA. The performance of GA-LM has been compared with that of conventional Back-Propagation (BP) algorithm and Levenberg–Marquardt(LM) algorithm. The comparison indicates that the predicted DO values using GA-LM model are in good agreement with the measured data. It is demonstrated here that the model is capable of predicting DO accurately, and can offer stronger and better performance than conventional neural networks when used as a quick interpolation and extrapolation tool.

Heuristic-rule-based hybrid neural network for navigation of a mobile robot

Abstract: The current paper presents a novel technique for a mobile robot to navigate in a real-world dynamic environment. When an autonomous mobile robot navigates in an unknown environment it is required to plan a path based on the information gathered from sensors in order to avoid obstacles and reach a target. This research idea is related to the basis of human perception, by using heuristic information for the navigation of mobile robots in cluttered dynamic environments which provides a general, robust, safe, and optimized path. The heuristic-rule-based network (HRBN) consists of a simple algorithm which makes the predefined estimation function much smaller. The estimation function should be adequately defined for desired movement in the environment. A navigation system using the rule-based technique allows a mobile robot to travel in an environment about which the robot has no prior knowledge. This heuristic rule is applied in conjunction with an artificial neural network (ANN). The ANN is trained by back-propagation algorithms. The HRBN provides an optimum trajectory which increases the effectiveness of a mobile robot. In a multiple-robot environment, a Petri net model (PNM) is used to prevent inter-robot collision during navigation. A series of simulations and experiments is conducted using mobile robots to show the effectiveness of the proposed algorithm.

Double-Layered Hybrid Neural Network Approach for Solving Mixed Integer Quadratic Bilevel Problems

Abstract: In this paper we build a double-layered hybrid neural network method to solve mixed integer quadratic bilevel programming problems. Bilevel programming problems arise when one optimization problem, the upper problem, is constrained by another optimization, the lower problem. In this paper, mixed integer quadratic bilevel programming problem is transformed into a double-layered hybrid neural network. We propose an efficient method for solving bilevel programming problems which employs a double-layered hybrid neural network. A two-layered neural network is formulate by comprising a Hopfield network, genetic algorithm, and a Boltzmann machine in order to effectively and efficiently select the limited number of units from those available. The Hopfield network and genetic algorithm are employed in the upper layer to select the limited number of units, and the Boltzmann machine is employed in the lower layer to decide the optimal solution/units from the limited number of units selected by the upper layer.The proposed method leads the mixed integer quadratic bilevel programming problem to a global optimal solution. To illustrate this approach, several numerical examples are solved and compared.

Neural assembly generation by selective connection weight updating

Abstract: In this paper, a neural network model, which learns symbols is introduced. VSF-Network(Vibration Synchronizing Function Network) is a hybrid neural network combining a chaos neural network with a hierarchical network. It has an ability for a incremental learning of patters by abstracting input data. VSF-Network finds unknown elements in data based on clusters generated by chaos neurons. VSF-Network generates sub-networks while it learns new patterns based on the information about the clusters. When a combination of learned patterns is presented, VSF-network recognizes them by combining its sub-networks. In this paper, an incremental learning model to examine the dynamics of VSF-network is introduced. The performances of the incremental learning by VSF-network are shown through the two tasks and the discussion about the combination form of the sub-networks generated by VSF-Network.

Novel Network Intrusion Detection System using Hybrid Neural Network (Hopfield and Kohonen SOM with Conscience Function)

Abstract: Summary Intrusion detection technology is an effective approach to dealing with the problems of network security. In this paper, it presents an intrusion detection model based on hybrid neural network and SVM. The key idea is to aim at taking advantage of classification abilities of neural network for unknown attacks and the expertbased system for the known attacks. We employ data from the third international knowledge discovery and data mining tools competition (KDDcup’99) to train and test the feasibility of our proposed neural network component. According to the results of our experiment, our model achieves 97.2 percent detection rate for DOS and Probing intrusions, and less than 0.04 percent false alarm rate. Expert system can detect R2L and U2R intrusions more accurately than neural network. Therefore, Hybrid model will improve the performance to detect intrusions.

Hybrid Neural Network Architecture for On-Line Learning

Abstract: Approaches to machine intelligence based on brain models use neural networks for generalization but they do so as signal processing black boxes. In reality, the brain consists of many modules that operate in parallel at different levels. In this paper we propose a more realistic biologically inspired hybrid neural network architecture that uses two kinds of neural networks simultaneously to consider short-term and long-term characteristics of the signal. The first of these networks quickly adapts to new modes of operation whereas the second one provides more accurate learning within a specific mode. We call these networks the surfacing and deep learning agents and show that this hybrid architecture performs complementary functions that improve the overall learning. The performance of the hybrid architecture has been compared with that of back-propagation perceptrons and the CC and FC networks for chaotic time-series prediction, the CATS benchmark test, and smooth function approximation. It is shown that the proposed architecture provides a superior performance based on the RMS error criterion.

On Bangla character recognition

Abstract: A recent surge of interest is to recognize Bangla characters. Bangla characters represent complex, multidimensional and meaningful visual information and developing a computational model for Bangla character recognition is a challenging job. This research presents a hybrid neural network solution for Bangla character recognition which combines local image sampling and artificial neural network. The method is based on BAM for dimensional reduction and multi-layer perception with backpropagation algorithm has been used for training the network. It has been found from practical observations that the number of iterations required to train the network is enormous. The capability of recognition of a neural network increases with increasing the training accuracy. For this process each character is converted to a designated M×N feature matrix. These feature matrices of characters are then fed into the neural network as input patterns .The neural network is trained with the set of input patterns of the digits to acquire separate knowledge corresponding to each Bangla character. In order to justify the effectiveness of the system, different test patterns of the characters are used to verify the system. Experimental results demonstrate that the system is capable of recognizing Bangla characters with 98% accuracy.

Hybrid neural network and case based reasoning system for Web user behavior clustering and classification

Abstract: In this paper we present Casep2: a hybrid neuro-symbolic system combining case-based reasoning (CBR) and artificial neural networks that aims at clustering and classifying users' behavior in an e-commerce site. A user behavior is represented by a sequence of visited web pages, in a session. Each registered behavior is associated to one of the following classes: buyer or non-buyer. Our goal is to provide a system that mines the web site access log in order to predict the class of an on-going user navigation. One major challenge to face is to provide scalable algorithms that can handle efficiently the large amount of data to learn from. Predictions should be made in real-time, during the current navigation. In addition, raw data has a sequential nature and are very noisy. In the proposed system, two original neural networks, named M-SOM-ART networks, are applied: one to implement the retrieval phase of a CBR cycle, and the second to implement the reuse phase. This hybrid scheme allows to ensure incremental learning as well as efficient treatment of large-scale sequential data. Experiments on real log data of an e-commerce site show the relevancy of the proposed approach.

Forecasting Models of Tool use in Different Intervals of Time

Abstract: In the paper the forecas ng models of tool use in di erent intervals of me were presented. The models were worked out by the use of hybrid neural networks in the form of: linear neural network (L) – mul -layer networks with error backpropaga on (MLP), L network – Radial Basis Func on network (RBF), MLP network – RBF network and L network – MLP network – RBF network. The comparison of these models was executed. The e ec veness of forecas ng of tool use in di erent me intervals is the measure of model evalua on. These models are used at the design stage of manufacturing process with the aim to plan produc on and prevent stands ll due to lack of tools, and special tools in par cular. The created models were tested on real data from an enterprise.

A Method of Hybrid Neural Network Adaptive Control for Flight Control System

Abstract: It is difficult to establish accurate models for complex flight control systems, but neural network has arbitrary nonlinear approximation ability. In order to overcome modeling errors and disturbances, a method of hybrid flight control is proposed. Firstly, inverse model of the object is identified online through neural networks and the feedback linearization control system is reached. And then circle theorem is used to design linear robust controller to control the state variables follow the input. A dynamic longitudinal model of a high-performance aircraft is considered to demonstrate the effectiveness of the proposed control scheme. Simulation results show designed controllers are highly adaptive and anti-interference ability.

Hybrid Neural Network Prediction Model of Hydraulic Valve Characteristics Under the Affection of Multiple Geometric Factors Effected

Abstract: Hydraulic valve system is a complex system with multiple characteristics affected by multiple geometric elements It will be essentially important to the manufacture process to establish the prediction model of the system characteristics by using the geometric elements and achieve the goal of prediction On the basis of synthesizing the features of the back propagation (BP) neural network and RBF neural network,a prediction model which is a new hybrid neural network based on the BP neural network and radial basis function (RBF) neural network is presented And the hybrid neural network is trained by using data measured from actual production The calculation results show that the hybrid neural network prediction model can improve the prediction accuracy of a single neural network model,and reach an average relative error of 0046 1 Therefore the proposed hybrid neural network can well satisfy the requirement of predicting the hydraulic valve characteristics