Showing papers on "Hybrid neural network published in 2016"

Learning Recursive Filters for Low-Level Vision via a Hybrid Neural Network

Abstract: In this paper, we consider numerous low-level vision problems (e.g., edge-preserving filtering and denoising) as recursive image filtering via a hybrid neural network. The network contains several spatially variant recurrent neural networks (RNN) as equivalents of a group of distinct recursive filters for each pixel, and a deep convolutional neural network (CNN) that learns the weights of RNNs. The deep CNN can learn regulations of recurrent propagation for various tasks and effectively guides recurrent propagation over an entire image. The proposed model does not need a large number of convolutional channels nor big kernels to learn features for low-level vision filters. It is significantly smaller and faster in comparison with a deep CNN based image filter. Experimental results show that many low-level vision tasks can be effectively learned and carried out in real-time by the proposed algorithm.

A Hybrid Double Feedforward Neural Network for Suspended Sediment Load Estimation

Abstract: Estimation of suspended sediment loads (SSL) in rivers is an important issue in water resources management and planning. This study proposes a hybrid double feedforward neural network (HDFNN) model for daily SSL estimation, by combining fuzzy pattern-recognition and continuity equation into a structure of double neural networks. A comparison is performed between HDFNN, multi-layer feedforward neural network (MFNN), double parallel feedforward neural network (DPFNN) and hybrid feedforward neural network (HFNN) models. Based on a case study on the Muddy Creek in Montana of USA, it is found that the HDFNN model is strongly superior to the other three benchmarking models in terms of root mean squared error (RMSE) and Nash-Sutcliffe efficiency coefficient (NSEC). HDFNN model demonstrates the best generalization and estimation ability due to its configuration and capability of physically dealing with different inputs. The peak value of SSL is closely estimated by the HDFNN model as well. The performances of HDFNN model in low and medium loads are satisfactory when investigated by partitioning analysis. Thus, the HDFNN is appropriate for modeling the sediment transport process with nonlinear, fuzzy and time-varying characteristics. It explores a practical alternative for use and can be recommended as an efficient estimation model for SSL.

Application of deep learning to the problem of vehicle detection in UAV images

Abstract: Detecting small objects such as vehicles in aerial images is a complicated problem, because it is difficult or impossible to find a suitable feature space to solve the problem for small objects. The aim of this work is to develop a system capable in real-time to solve the challenge of detection the vehicles. Deep convolutional networks can automatically extract rich features from the training sample and achieve good performance on a variety of data. In this paper, we present the adapted hybrid neural network (HDNN) in which the last layers are divided into several blocks of variable size so that the network could extract features of different scales. Experimental results show that HDNN, which was offered, exceeds the results of other conventional transport methods of detection.

Knowledge Enhanced Hybrid Neural Network for Text Matching

Abstract: Long text brings a big challenge to semantic matching due to their complicated semantic and syntactic structures. To tackle the challenge, we consider using prior knowledge to help identify useful information and filter out noise to matching in long text. To this end, we propose a knowledge enhanced hybrid neural network (KEHNN). The model fuses prior knowledge into word representations by knowledge gates and establishes three matching channels with words, sequential structures of sentences given by Gated Recurrent Units (GRU), and knowledge enhanced representations. The three channels are processed by a convolutional neural network to generate high level features for matching, and the features are synthesized as a matching score by a multilayer perceptron. The model extends the existing methods by conducting matching on words, local structures of sentences, and global context of sentences. Evaluation results from extensive experiments on public data sets for question answering and conversation show that KEHNN can significantly outperform the-state-of-the-art matching models and particularly improve the performance on pairs with long text.

Knowledge Enhanced Hybrid Neural Network for Text Matching

Abstract: Long text brings a big challenge to neural network based text matching approaches due to their complicated structures. To tackle the challenge, we propose a knowledge enhanced hybrid neural network (KEHNN) that leverages prior knowledge to identify useful information and filter out noise in long text and performs matching from multiple perspectives. The model fuses prior knowledge into word representations by knowledge gates and establishes three matching channels with words, sequential structures of text given by Gated Recurrent Units (GRUs), and knowledge enhanced representations. The three channels are processed by a convolutional neural network to generate high level features for matching, and the features are synthesized as a matching score by a multilayer perceptron. In this paper, we focus on exploring the use of taxonomy knowledge for text matching. Evaluation results from extensive experiments on public data sets of question answering and conversation show that KEHNN can significantly outperform state-of-the-art matching models and particularly improve matching accuracy on pairs with long text.

Hybrid neural network approach for predicting signal propagation loss in urban microcells

Abstract: In the last few years, the need for connectivity far and wide, coupled with the continuous increase in the number of cellular network subscribers globally, has stirred the development and evolution of diverse cellular communication standards. This in turn has led to speedy installations of base station transmitters, thus making the process of planning and fine-tuning the location of these BS transmitters very difficult. To plan and optimize mobile cellular networks for acceptable level of service coverage and quality at the mobile station terminals, radio network engineers rely on propagation loss prediction models. This research work investigates the application of a neural hybridized model for field signal strength attenuation prediction. The hybridized model combines a conventional Log-distance model and an adaptive neural network model. The adaptive neural model employs a multilayer Levenberg Marquardt back propagation algorithm to reimburse for the prediction errors obtained by means of using only the conventional model in urban microcellular environment. After applying a number of first order statistical indicators such standard deviation and root mean square error for a comprehensive performance evaluation, the hybrid — based algorithm provides more accurate prediction results with measured values compared to the conventional approach. The computationally effective prediction technique of the hybrid based neural network model can be used for standardization and enhancement of the conventional field strength propagation loss prediction methods.

Prediction of Hydraulic Conductivity of Soil Bentonite Mixture Using Hybrid-ANN Approach

Abstract: Due to its low hydraulic conductivity compacted soil-bentonite mixture is widely used as a barrier material at waste disposal site. The experimental determination of hydraulic conductivity of soil-bentonite mixture, which depends on the various physical and chemical and mineralogical factors, requires expensive and time consuming setup. Thus, a hybrid neural network model (combining genetic algorithm with neural network) is presented here as a complementary tool to model hydraulic conductivity of soil-bentonite mixture. The prediction capability of the model has been found to be satisfactory. The developed model yielded correlation coefficients of 0.98 and 0.97 for training and testing data sets, respectively. The proposed model was compared with conventional neural network models by using different statistical indicators such as Nash-Sutcliffe model efficiency and discrepancy ratio with standard deviation. It was found that the predictions obtained from developed model agreed well with experimental observations. Identification of important parameters and ranking their order of influence on hydraulic conductivity has been discussed by using input significance test.

Two-dimensional magnetic modeling of ferromagnetic materials by using a neural networks based hybrid approach

Abstract: This paper presents a hybrid neural network approach to model magnetic hysteresis at macro-magnetic scale. That approach aims to be coupled together with numerical treatments of magnetic hysteresis such as FEM numerical solvers of the Maxwell's equations in time domain, as in case of the non-linear dynamic analysis of electrical machines, and other similar devices, allowing a complete computer simulation with acceptable run times. The proposed Hybrid Neural System consists of four inputs representing the magnetic induction and magnetic field components at each time step and it is trained by 2D and scalar measurements performed on the magnetic material to be modeled. The magnetic induction B is assumed as entry point and the output of the Hybrid Neural System returns the predicted value of the field H at the same time step. Within the Hybrid Neural System, a suitably trained neural network is used for predicting the hysteretic behavior of the material to be modeled. Validations with experimental tests and simulations for symmetric, non-symmetric and minor loops are presented.

Hybrid Neural Network Model for Metocean Data Analysis

Abstract: Metocean time-series data is generally classified as highly chaotic thus making the analysis process tedious. The main aim of forecasting Metocean data is to obtain an effective solution for offshore engineering projects, such analysis of environmental conditions is vital to the choices made during planning and operational stage which must be efficient and robust. This paper presents an empirical analysis of Metocean time-series using a hybrid neural network model by performing multi-step-ahead forecasts. The proposed hybrid model is trained using a gauss approximated Bayesian regulation algorithm. Performance analysis based on error metrics shows that proposed hybrid model provides better multi-step-ahead forecasts as in comparison to previously used models.

Hybrid Neural Network and Genetic Algorithm for off-Lexicon Online Arabic Handwriting Recognition

Abstract: In this paper we propose the hybridization of neural networks and genetic algorithm for online Arabic handwriting recognition. The used method consists in decomposing the input signal into continuous parts called graphemes based on Beta-Elliptical model and baseline detection. The segmented graphemes are then described according to their position in the pseudo-word by a combination of geometric features modeling their trajectory shape and provided in the input of the neural networks used for graphemes class recognition. Finally, a genetic algorithm is used to generate the characters code corresponding to the obtained chain of recognized graphemes code by applying the genetic search process: selection, crossover and mutation. The developed system is evaluated using an Arabic words dataset extracted from the ADAB Database.

Design of slot loaded proximity coupled microstrip antennas using knowledge based neural networks

Abstract: In this paper, a knowledge based hybrid neural network (KBHNN) is utilized for designing of different slotted proximity coupled microstrip antennas. The slot loaded antennas can be designed from 1 to 6 GHz frequency ranges. By using this model, accuracy is found to be really beneficial, even if the required number of training data has been brought down to half. This method requires less time and scales down the complexities of the design processes. The solutions obtained by this neural approach are compared with the CST simulation results. The results of the KBHNN method are in good accord with the simulated values.

Hybrid Neural Network mixed with Random Forests and Perlin noise

Abstract: Data mixed up with discrete and continuous features makes negative affect to the classification for existing models which discretizes the continuous features or even without any treatment is not able to deal with it. The noise which is not subject to Gaussian distribution also severely affected the result of classification. In this paper, a deep neural network mixed with random forests, Stacked Denoising Autoencoder and Multilayer Perception is proposed to improve the classification effect of mixed features input. Furthermore, to improve pre-training effect of the Stacked Denoising Autoencoder, Perlin noise is added to the Gaussian Noise in noise layer. A certain promotion of this method has been proved through the experiments with the crime data between 2003 and 2015 in San Francisco, and comparisons by 10-fold cross-validation with other common methods of classification shows that this method has advantages on data mixed up with discrete and continuous features.

Intelligent Soft Computing on Forex: Exchange Rates Forecasting with Hybrid Radial Basis Neural Network.

Abstract: This paper deals with application of quantitative soft computing prediction models into financial area as reliable and accurate prediction models can be very helpful in management decision-making process. The authors suggest a new hybrid neural network which is a combination of the standard RBF neural network, a genetic algorithm, and a moving average. The moving average is supposed to enhance the outputs of the network using the error part of the original neural network. Authors test the suggested model on high-frequency time series data of USD/CAD and examine the ability to forecast exchange rate values for the horizon of one day. To determine the forecasting efficiency, they perform a comparative statistical out-of-sample analysis of the tested model with autoregressive models and the standard neural network. They also incorporate genetic algorithm as an optimizing technique for adapting parameters of ANN which is then compared with standard backpropagation and backpropagation combined with K-means clustering algorithm. Finally, the authors find out that their suggested hybrid neural network is able to produce more accurate forecasts than the standard models and can be helpful in eliminating the risk of making the bad decision in decision-making process.

Learning Symbols by Neural Network

Abstract: VSF−Network is a neural network model that learns dynamical patterns. It is hybrid neural network combining a chaos neural network and a hierarchical neural network. The hierarchical neural network learns patterns and the chaos neural network monitors behavior of neurons in the hierarchical neural network. In this paper, two theoretical backgrounds of VSF−Network are introduced. An incremental learning framework using chaos neural networks is introduced. The monitoring by chaos neural network is based on clusters generated by synchronous vibration. Using the monitoring results, redundant neurons in the hierarchical neural network are found and they are used for learning of new patters. The second background is about the pattern recognition by combining learned patterns. This is explained by code words expression used in multi-level discrimination. Through an experiment, both the incremental learning capability and the pattern recognition are shown.

Hybrid neural network algorithm-based performance assessment method used for complex industrial product

Abstract: The invention discloses a hybrid neural network (HNN) algorithm-based performance assessment method used for a complex industrial product The method comprises the steps of firstly determining HNN structure parameters; secondly initializing a connection weight of a neuron of each layer and a characteristic point of a membership function, and setting an error limit value, an iterative frequency, a learning rate and a momentum factor; thirdly performing quantification and normalization on input fuzzy sample data, and performing normalization on a quantitative numerical value; fourthly calculating a learning error derivative of the neuron of each layer in repeated iteration, correcting the connection weight, and adjusting the characteristic point of the membership function by adopting a gradient descent method; and finally performing repeated iteration until a set error is reached, and giving out a quantitative performance prediction result of the complex industrial product through an HNN algorithm The HNN algorithm provides a solution with high prediction accuracy for the problem in complex industrial performance assessment based on qualitative, quantitative and qualitative-quantitative combined data input and possibly different dimension numbers of input data items in actual conditions

A hybrid neural network using ICA and CGA for skin detection in RGB images

Abstract: Skin color recognition is a useful and popular method in human-computer interaction and also in analyzing the content. In addition, the application programs for recognizing and detecting human body parts, faces, naked people, and retrieving individuals in multimedia databases all make use of skin recognition. Thus, finding a suitable method in order to segment the pixels of an image into different groups such as skin can be very important. Imperialist competitive algorithm (ICA) is a recently introduced evolutionary algorithm that showed a promising performance in some of the optimization problems. In this article, first the combined ICA-ANN, continuous genetic algorithm (CGA) and gradient descent algorithm were proposed and their performance was tested on images in RGB color spaces. In the proposed algorithms, a multilayer perceptron neural network manages the problem's constraints, and ICA and genetic algorithms search to calculate the best response than the gradient descent algorithm. The proposed skin classification algorithms perform directly on the RGB color space. The results clearly indicate that the proposed algorithm significantly improves the performance of an MLP neural network.

Digital-analog hybrid neural network chip architecture

Abstract: The invention discloses a digital-analog hybrid neural network chip architecture. The architecture comprises a two-dimensional SRAM module, an analog synaptic circuit, a nerve cell circuit, an AER communication module, and a master control digital unit. The two-dimensional SRAM module is taken as a storage unit of neural network connection relation and a synaptic weight value. The analog synaptic circuit and the nerve cell circuit respectively consist of an MOSFET circuit working in a subthreshold section. The AER communication module serves as the input and output interfaces of a chip, and employs an AER protocol for communication. All control circuits in the architecture are synchronous digital circuits. The architecture is low in power consumption, is high in degree of parallelism, and can achieve a neural network algorithm in a reasonable chip area, wherein the neural network algorithm is more complex in nerve cell functions, is larger in network scale, and is more flexible in connection.

Навчання модулів нейронної мережі

Abstract: Currently, there exists a huge number of neural networks of different classes, each with itsown advantages and disadvantage. However, there aren’t a lot of focus on hybrid neural networks, basedon the combination of knowт topologies of neural networks. Modular organization principle seems to bevery promising, however principles of its module creation isn’t known and needs further research. Thepresent study, therefore, proposes some methods of hybrid neural network module creation and theirlearning algorithms

Detect the emotions of the public based on cascade neural network model

Abstract: Along with the development of social network, more and more people know the world by reading news. The problem about what kind of emotion is inspired when people read news is very worthy of discussion. This paper will mix Deep Belief Networks (DBN) model and Support Vector Machine (SVM) to a hybrid neural network model by using the Contrast Divergence (CD) algorithm to estimate the weights when training a generating model, ensure that each layer of the Restricted Boltzmann Machine (RBM) mapping the features of the inputs to the best. At the same time, we cascade the last layer of DBN and a SVM classifier to adjust judging performance. And a set of tags will be attached to the top (Associative Memory), through a process of parameter tuning, learn the identifying weights to obtain a network for the task of text classification. The experimental results show that the hybrid neural network model works better than the traditional text categorization method based on simple characteristics (such as CHI), and it is more suitable for extracting text semantic characteristics.

Day-Ahead Electricity Spike Price Forecasting Using a Hybrid Neural Network-Based Method

Abstract: A hybrid neural network-based method is presented to predict day-ahead electricity spike prices in a deregulated electricity market. First, prediction of day-ahead electricity prices is carried out by a neural network along with pre-processing data mining techniques. Second, a classifier is used to separate the forecasted prices into normal and spike prices. Third, a second neural network is trained over spike hours with selected features and is used to forecast day-ahead spike prices. Forecasted spike and normal prices are combined to produce the complete day-ahead hourly electricity price forecasting. Numerical experiments demonstrate that the proposed method can significantly improve the forecasting accuracy.

Гибридные модулярные нейронные сети

Abstract: In this paper considered hybrid neural networks for time series forecasting and methods based on fuzzy models, such as ANFIS models and we also suggest a hybrid neural network model based on modular neural networks.

Low-rate ddos attack detection using hybrid neural network

Abstract: Представлены результаты разработки метода обнаружения низкоинтенсивных атак типа «отказ в обслуживании» на http-сервисы. Используется модель представления низкоинтенсивной атаки в виде упорядоченного по времени ряда событий с аддитивным наложением атакующего воздействия и легального трафика. Такое представление позволяет задействовать математический аппарат обработки сигналов, включая методы распознавания образов. Задача разработки метода обнаружения низкоинтенсивных атак сформулирована как последовательное решение задач выделения гомогенных групп (patterns) временного ряда на основании моделей распознавания образов (pattern recognition), и последующего построения для каждой группы отдельной модели прогнозирования. С учетом контекста решаемой задачи (требования на высокую точность классификации, скорость формирования моделей и скорость классификации) определено, что наиболее перспективным направлением решения является использование комбинированных нейросетевых моделей, выполняющих на первом этапе кластеризацию, а затем прогнозирование временного ряда внутри установленного кластера. Непосредственно для обнаружения атаки необходимо выявлять факт периодического появления однотипного набора пакетов во входящем трафике и затем определять принадлежность этого набора к определённому классу (нормальному или аномальному). Порядок следования пакетов при этом ведущей роли не играет, временная информация учитывается при разбиении входящего трафика на окна. Метод включает следующие шаги: 1) для каждого защищаемого сервиса построить отдельную гибридную ИНС; 2) Для отдельного сервиса получить набор пакетов, число которых определено величиной окна (экспериментально установленное значение); 3) сформировать вектора для шага снижения размерности (самоорганизующейся карты); 4) снизить размерность входных данных путём кластеризации вектора при помощи SOM; 5) построить вектора для MLP, в которых каждый компонент соответствует номеру кластера, которому принадлежит пакет. Т.о. входным вектором будет набор кластеризованных пакетов, сохраняющий информацию о порядке (последовательности) поступления. Для пакетов уже определена принадлежность к определённому типу; 6) проанализировать вектора на MLP, классифицировать выявленные в трафике набор, разделив их на два класса – атака или норма. Обнаружение атак; низкоинтенсивные DDoS-атаки; отказ в обслуживании; искусственная нейронная сеть; гибридная нейронная сеть; безопасность вычислительных сетей.

Hybrid neural network training algorithm for spectrum sensing in cognitive radio networks

Abstract: Spectrum sensing in the cognitive radio networks has emerged as a highly gifted technique which has riveted the eagle eyes of the enthusiastic experimenters for the last few years. In the novel method, the Levenberg-Marquardt-based neural network is integrated with the most modern optimisation technique, known as the gravitational search algorithm with an eye on perking up the proficiency in sensing of the channel. The user data is communicated by means of the redundant or vacant channels presently in the system by the deft deployment of the Levenberg-Marquardt-based neural network GS-LM approach, where the channel is located; in accordance with the channel state forecast outcomes. The relative evaluation is performed by assessing and contrasting the outcomes of the innovative approach to those of the HMM, LM-based NN and arbitrary technique. The maximum SU and SUimp values attained by the novel method are approximately 0.58 and 0.41 correspondingly.

Hybrid Neural network Training Algorithm For Spectrum Sensing In Cognitive Radio Networks

Abstract: Spectrum sensing in the cognitive radio networks has emerged as a highly gifted technique which has riveted the eagle eyes of the enthusiastic experimenters for the last few years. In the novel method, the Levenberg-Marquardt-based neural network is integrated with the most modern optimisation technique, known as the gravitational search algorithm with an eye on perking up the proficiency in sensing of the channel. The user data is communicated by means of the redundant or vacant channels presently in the system by the deft deployment of the Levenberg-Marquardt-based neural network GS-LM approach, where the channel is located; in accordance with the channel state forecast outcomes. The relative evaluation is performed by assessing and contrasting the outcomes of the innovative approach to those of the HMM, LM-based NN and arbitrary technique. The maximum SU and SUimp values attained by the novel method are approximately 0.58 and 0.41 correspondingly.

Detection of Human Faces Using Neural Networks

Abstract: Human face detection is a key technology in machine vision applications including human recognition, access control, security surveillance and so on. This research proposes a precise scheme for human face detection using a hybrid neural network. The system is based on visual information of the face image sequences and is commenced with estimation of the skin area depending on color components. In this paper we have considered HSV and YCbCr color space to extract the visual features. These features are used to train the hybrid network consisting of a bidirectional associative memory BAM and a back propagation neural network BPNN. The BAM is used for dimensional reduction and the multi-layer BPNN is used for training the facial color features. Our system provides superior performance comparable to the existing methods in terms of both accuracy and computational efficiency. The low computation time required for face detection makes it suitable to be employed in real time applications.

A hybrid neural network method for simulating spatial variation in earthquake ground motion

Abstract: Different excitations for supports should be considered for the analysis of long-span structures. The excitation of each support has time delay and spatial variation relative to other support excitations. The present study aims to propose a new method for simulating accelerograms for various distances considering spatial variation of earthquake records. The accelerograms are simulated based on response or design spectra using the learning capabilities of neural networks. In this method, the response spectrum, and the distance parameter (distance from fault rupture) are the input, and the corresponding accelerograms are the output of the network. There are three stages involved in this study. In the first stage, a replicator neural network is used as a data compressor to increase capability of the simulation. In the second stage, a radial basis function neural network is employed to generate a compressed accelerogram for a certain distance and a response spectrum. In the third stage, the compressed...

Feature Based Hybrid Neural Network for Hand Gesture Recognition

Abstract: This paper presents using neural network as a method for classifying hand gesture effectiveness in home appliances or serious games. Neural learning from imbalanced data has some difficulties, but, we presents Feature based Hybrid Neural Network(FHNN) that simple calculation can extract data distribution feature and add it to input layer for Neural learning. As data distribution feature, we used particular point of gestures to get approximate classification and extrema of gesture trajectory. The experimental results show that FHNN can outperform the compared methods.