Weight Function Neural Networks Research Articles

Privacy Preserving Over Big Data Through VSSFA and MapReduce Framework in Cloud Environment

The main intention of this paper is develop a technique for privacy preserving-aware over big data in clouds using variation step size firefly algorithm and MapReduce framework. This paper consists of two phases such as MapReduce phase and evaluation phase. In MapReduce phase, at first we split the big data into a number of maps. After that, we apply the convolution process to the dataset and create a new kernel matrix. Once the convolution process is over, the privacy-persevering framework over big data in cloud systems is performed based on the evaluation phase. In evaluation module, the neural network is trained based on the optimal weight and radial basis function neural network algorithm, which is improving the utility of the privacy data. After that, a MapReduce framework is to protect the private information. Finally, the reduced privacy data are stored in the computer service provider. In this big data experiment, we used census income KDD dataset to evaluate the result.

Environmental Quality Assessment Based on a New Kind of Neural Network Using Rational Spline Weight Functions

As we all known, artificial neural network can be used in the process of environmental quality assessment. To improve the accuracy and science of assessment, a method of environmental quality assessment is presented in this paper, which is based on spline weight function (SWF) neural networks. The weigh functions of the neural network are composed of rational spline functions with cubic numerator and linear denominator (3/1 rational SWF). The simulation results show that, compared with the conventional BP neural networks, this method can get very high precision and accuracy. This case demonstrates that SWF neural networks can offer a very prospective tool for environmental quality assessment.

Sleep Sensor Temperature Prediction Based on Neural Network Training Algorithm with Rational Spline Weight Functions

In wireless sensor network, it is necessary to make effective prediction of sensor node’s data during its sleep period. In this paper a model of rational cubic spline weight function (SWF) neural network with linear denominator was established for sensor node’s temperature prediction. This kind of rational spline function is denoted by 3/1 rational splines. Then we trained and tested the network, the simulation results showed that, compared to the traditional BP neural network, the training speed is higher and the error is smaller. Therefore the prediction model can effectively predict the sensor’s temperature.

Passenger Flow Analysis in Subway Using a Kind of Neural Network

This paper aims to analyze passenger flow in subway based on a kind of rational spline weight function neural network, in which the numerator of the spline is a cubic polynomial and the denominator of the spline is a quadratic polynomial, and this kind spline is denoted by 3/2 rational splines. There are many factors affecting the passenger flow. Combined the main influential factors with the self-learning method of neural network, we establish the neural network model of passenger flow in subway. This paper introduces the spline weight function neural network and the passenger flow model based on this neural network. Finally MATLAB simulation verifies that the 3/2 rational spline weight function neural network can be applied to analyze the passenger flow in subway with high accuracy.

Error Analysis of Neural Network with Rational Spline Weight Function

Training neural network by spline weight function (SWF) has overcomed many defects of traditional neural networks (such as local minima, slow convergence and so on). It becomes more important because of its simply topological structure, fast learning speed and high accuracy. To generalize the SWF algorithm, this paper introduces a kind of rational spline weight function neural network and analyzes the performance of approximation for this neural network.

Research of the First Kind of Padé Weight Function Neural Networks

Traditional algorithms of training neural networks have some weakness. A new approach using weight function neural networks was proposed in recent years, which overcomes the shortcomings of traditional neural networks. This paper takes Padé approximation as weight functions and we get the method of how to find the Padé weight functions with given samples. Then complexity is also analysed in detailed in this paper. Finally simulation experiments are done to verify the results.

Complexity Research on Second Class Orthogonal Weight Function Neural Networks

Weight function neural network is a new kind of neural network developed in recent years, which has many advantages, such as finding globe minima directly, good performance of generalization, extracting some useful information inherent in the problems and so on. Time complexity is an important measure of algorithm. This paper studies the complexity of neural network using second class orthogonal weight functions. The results indicate that the neural network has a linear relationship with the dimensions of input layer and output layer, an O (n3) relationship with the number of samples. Finally gives some simulation experiments for time complexity.

Time Complexity Research on the Second Type Neural Networks of Padé Weight Functions

Time complexity is an important measure of algorithm. The main purpose of this paper is to research the time complexity of the second category of Padé weight function neural network and find out the factors which affect its time complexity. In this paper, firstly, the second category of Padé weight function neural network algorithm is introduced. Then through the analysis of the key steps of the algorithm, the time complexity is given. After MATLAB simulation, the experimental results verify the theoretical analysis of the results. Therefore, its time complexity is related to input dimension, output dimension and the number of training samples.

Speech Recognition Based on Weight Function Neural Networks

Weight function neural network is a new kind of neural network developed in recent years, which has many advantages, such as finding globe minima directly, good performance of generalization, extracting some useful information inherent in the problems and so on. In this article, we apply orthogonal weight function neural network to the speech recognition. The result indicates that the weight function neural network has a good efficiency in speech recognition.

A New Theory of Infinite Series and its Applications in Training Weight Function Neural Networks

A new theory of infinite series is proposed in this paper, some new important theorems for function expansion and infinite series are also proposed. Unlike Taylors expansion, the expansion generated by a function is not the form of polynomials. In general, the performance of convergence is much better than that obtained by Taylor's Series. The new important theorems lay the foundation for the new theory of infinite series and applications. To describe the performance of the new results obtained in this paper, an example given in this paper shows that the region of convergence is much larger than that of Taylors series. The new infinite series can keep some important properties of original functions. Weight function neural networks are also used to training feedforward neural networks based on the new theory proposed in this paper.

Chinese Word Segmentation Based on the First Kind of Spline Weight Function Neural Networks

With the continuous development of information technology, word segmentation technology becomes an important link in dealing with the increasing amount of information conveniently. Different from English word segmented by spaces, Chinese writing is continuous, and there is no space between words, this brings a lot of trouble to word segmentation. In this article, through the analysis of different property of words to get a code that can be used for training and combined it with the first kind of spline weight function neural network, then by training a large number of existing rules encoding to generate a study method that can divide the statement correctly.

Implementation of Mail Classification Using Neural Networks of the Second Type Spline Weight Functions

How to effectively filter out spam is a topic worthy of further study for the growing proliferation of spam. The main purpose of this paper is to apply a new neural network algorithm to the classification of spam. In this paper, we introduce a second type of spline weight function neural network algorithm, as well as e-mail feature extraction and vectorization, and then introduced the mail sorting process. Experiments show that it can get a relatively high accuracy and recall rate on the spam classification. Therefore, with this new algorithm, we can achieve better classification results.

Short-Term Traffic Flow Forecasting of Road Based on Spline Weight Function Neural Networks

Traditional short-term traffic flow forecasting of road usually based on back propagation neural network, which has a low prediction accuracy and convergence speed. This paper introduces a spline weight function neural networks which has a feature that the weight function can well reflect sample information after training, thus propose a short-term traffic flow forecasting method base on the spline weight function neural network, specify the network learning algorithm, and make a comparative tests bases on the actual data. The result proves that in short-term traffic flow forecasting, the spline weight function neural network is more effective than traditional methods.

Sensitivity of B-Splines Weight Function Neural Networks and its Application on Signal Recognition

The statistical sensitivity of training neural networks by B-splines weight functions and its applications for digital modulation signal recognition (DMSR) is discussed in this paper. By extracting some instantaneous parameters using the technology of DMSR, and using the sensitivity formula of B-splines weight function neural networks proposed in this paper, we show that the classifier of B-splines weight function neural networks has optimized architecture and high recognition rate.

Statistical Sensitivity Analysis for Training Feedforword Neural Networks by B-Splines Weight Functions

Based on the new architecture and algorithm of training neural networks by B-splines weight functions, the statistical sensitivity analysis for neural networks using B-splines weight functions is discussed in this paper. The sensitivity formula of B-splines weight function neural networks is derived. Finally, the correctness of theoretical sensitivity formula is verified by simulation.

Identification of Road Surface Power Spectrum Density Based on a New Cubic Spline Weight Neural Network

Aiming at the shortcomings appearing in the BP neural network method for obtaining the road surface power spectrum density, the paper applies a new cubic spline weight function neural network (CSWFNN) to the identification of road surface power spectrum density. First, a vibration model with four freedoms is established to simulate the actual vehicles. Subsequently, we deduce the nonlinear relationship between power spectrum density of the response and that of the excitation, and then acquire some input and output data. Next, a CSWFNN is built and trained until converges. In the end, the paper analyses the errors of CSWFNN in the identification of road surface power spectrum density. The results show that the proposed method in the paper successfully overcomes the deficiencies of BP neural network method, and consequently enhances the identification precision of road surface power spectrum density.