A Survey on Neural Network Models for Data Analysis

We explore how techniques that were developed to improve the training process of artificial neural networks can be used to speed up fuzzy clustering. The basic idea of our approach is to regard the difference between two consecutive steps of the alternating optimization scheme of fuzzy clustering as providing a gradient, which may be modified in the same way as the gradient of neural network backpropagation is modified in order to improve training. Our experimental results show that some methods actually lead to a considerable acceleration of the clustering process.

Signal Processing, 1994

Winner-take-all algorithms are commonly used techniques in clustering analysis. However, they have some problems ranging from clusters under utilization to the extended training time. Some solutions to these problems are addressed here. It is shown here that using the maximum-likelihood criterion instead of the Euclidean distance metric results in better clustering. The clusters are represented by a set of neuron each has a Gaussian receptive field. For these Gaussian neurons, the covariance matrices, in addition to the centers, are learned. The one-winner condition is relaxed by maximizing the likelihood function of the mixture density function of the samples. This produces larger likelihood values and more normally distributed clusters. A fast mixture likelihood clustering is provided for both batch and pattern learning modes. Convergence analysis and experimental results are also presented.

Cybernetics and Systems, 2006

A new approach, designed for clustering of arbitrary distributed patterns, is presented. This study is concerned with the use of a selforganizing neural network as a frame for data clustering. The nearest network nodes in feature space are treated as prototypes, assigned to the corresponding cluster. The rules for dead-node shifting and simple adjustment of coordinates of the active nodes are introduced. The performance of the proposed self-organizing neural network is examined on the benchmark synthetic and the real-world problem.

Introduction As long as knowledge is another name of power, many organizations give much importance to knowledge. When reaching this knowledge, they make use of the data in their databases. Data is important since it enable them to learn from the past and to predict future trends and behaviours (Larose, 2005). Today, most of the organizations use the data collected in their databases when taking strategic decisions (Singh and Chauhan 2009). The process of using the data to reach this knowledge consists of two steps as to collecting the data and analysing the data. In the beginning, organizations are faced with difficulties when collecting the data so they have not enough data in order to make suitable analysis (Kartalopoulos and Stamatios, 1996). In the long run, with the rapid computerization, organizations are able to store huge amount of data easily. But at this time, they are faced with another problem when analysing and

Expert Systems with Applications, 2012

The statistical properties of training, validation and test data play an important role in assuring optimal performance in artificial neural networks (ANNs). Researchers have proposed optimized data partitioning (ODP) and stratified data partitioning (SDP) methods to partition of input data into training, validation and test datasets. ODP methods based on genetic algorithm (GA) are computationally expensive as the random search space can be in the power of twenty or more for an average sized dataset. For SDP methods, clustering algorithms such as self organizing map (SOM) and fuzzy clustering (FC) are used to form strata. It is assumed that data points in any individual stratum are in close statistical agreement. Reported clustering algorithms are designed to form natural clusters. In the case of large multivariate datasets, some of these natural clusters can be big enough such that the furthest data vectors are statistically far away from the mean. Further, these algorithms are computationally expensive as well. We propose a custom design clustering algorithm (CDCA) to overcome these shortcomings. Comparisons are made using three benchmark case studies, one each from classification, function approximation and prediction domains. The proposed CDCA data partitioning method is evaluated in comparison with SOM, FC and GA based data partitioning methods. It is found that the CDCA data partitioning method not only perform well but also reduces the average CPU time.

IEEE Transactions on Neural Networks, 2000

Proceedings of the 1996 ACM symposium on Applied Computing - SAC '96, 1996

Conventional clustering techniques group input feature vectors in crisp classes. However, in practice the input feature vector may belong to more than one class. This is specially the case for pixels in a satellite image. Often, the area represented by a pixel may belong to more than one category, i. e. half the area may belong to water and remaining half may belong to land. In the situation such as this, we would like to use partial memberships. Fuzzy sets allow partial memberships. In this paper we have suggested an algorithm for clustering that combines neural networks with fuzzy logic. We have used the model successfully to analyze satellite images. The results are presented in the paper.

2016

There are plenty of technologies available to data mining practitioners, including Artificial Neural Networks, Regression, and Decision Trees. Data mining tools can forecast the future trends and activities to support the decision of Officials. Some data mining tools can also solve some traditional problems which consumed enough time, this is because that they can rapidly browse the entire database and find some useful information experts undetected. Neural network is a parallel processing network which made with simulating the intuitive thinking of human, the neural network in data mining was not optimistic, and the key reasons are that the nerve organs network has the problems of complex structure, poor interpretability and long training time. Nevertheless advantages such as high affordability to the noise data and low error rate, the consistently advancing and optimization of various network training algorithms, especially the continually advancing and improvement of varied netwo...

 Abstract— One of the greatest potential of applying intelligent multi agent systems is the support of machine learning in order to reflect the whole complexity of the real and virtual world. However, a critical deficiency is a gap between two applicable streams of intelligent multi agent technology and learning models. In order to solve this problem, we developed a framework of intelligent multi agent based on the efficient feed-forward neural network clustering method. A feed-forward neural network is a software version of the brain and a popular tool for statistical decision making. The framework is applicable to different domains successfully and for the potential case study, the clinical domain and the breast cancer database from the University of Malaya Medical Center is considered to predict the survival time. (Abstract)

Journal of Computing in Civil Engineering, 2004