Self-organizing Kohonen Research Articles

Application of Clustering Techniques Using Prioritized Variables in Regional Flood Frequency Analysis—Case Study of Mahanadi Basin

The selection of suitable site characteristics and the number of clusters play an important role for finding homogeneous regions in regional flood frequency analysis. The present study investigates the partition of the Mahanadi basin into homogeneous regions by applying different clustering techniques by using fewer but influential variables. As such, the entire basin is not hydrometeorologically homogeneous. Principal component analysis has been initiated in finding appropriate site characteristics (variables) as per priority. Out of seven variables, four variables are selected on priority. Possible numbers of cluster are found by applying the Kohonen self-organization map and Andrews plot. Other clustering techniques, such as hierarchical clustering fuzzy C-mean (FCM) and K -mean, are applied on prioritized variables to verify the result of clustering. The intercomparison of clustering techniques gives the optimum number of sites to be placed in a particular cluster. The sites clustered as per FCM give ...

Applications of neural networks in training science

Training science views itself as an integrated and applied science, developing practical measures founded on scientific method. Therefore, it demands consideration of a wide spectrum of approaches and methods. Especially in the field of competitive sports, research questions are usually located in complex environments, so that mainly field studies are drawn upon to obtain broad external validity. Here, the interrelations between different variables or variable sets are mostly of a nonlinear character. In these cases, methods like neural networks, e.g., the pattern recognizing methods of Self-Organizing Kohonen Feature Maps or similar instruments to identify interactions might be successfully applied to analyze data. Following on from a classification of data analysis methods in training-science research, the aim of the contribution is to give examples of varied sports in which network approaches can be effectually used in training science. First, two examples are given in which neural networks are employed for pattern recognition. While one investigation deals with the detection of sporting talent in swimming, the other is located in game sports research, identifying tactical patterns in team handball. The third and last example shows how an artificial neural network can be used to predict competitive performance in swimming.

Adaptive Shadow and Highlight Invariant Colour Segmentation for Traffic Sign Recognition Based on Kohonen SOM

This paper describes an intelligent algorithm for traffic sign recognition which converges quickly, is accurate in its segmentation and adaptive in its behaviour. The proposed approach can segment images of traffic signs in different lighting and environmental conditions and in different countries. It is based on using Kohonen's Self-Organizing Maps (SOM) as a clustering tool and it is developed for Intelligent Vehicle applications. The current approach does not need any prior training. Instead, a slight portion, which is about 1% of the image under investigation, is used for training. This is a key issue to ensure fast convergence and high adaptability. The current approach was tested by using 442 images which were collected under different environmental conditions and from different countries. The proposed approach shows promising results; good improvement of 73% is observed in faded traffic sign images compared with 53.3% using the traditional algorithm. The adaptability of the system is evident from the segmentation of the traffic sign images from various countries where the result is 96% for the nine countries included in the test.

Matrix Neuro-Fuzzy Self-Organizing Clustering Network

Matrix Neuro-Fuzzy Self-Organizing Clustering NetworkIn this article the problem of clustering massive data sets, which are represented in the matrix form, is considered. The article represents the 2-D self-organizing Kohonen map and its self-learning algorithms based on the winner-take-all (WTA) and winner-take-more (WTM) rules with Gaussian and Epanechnikov functions as the fuzzy membership functions, and without the winner. The fuzzy inference for processing data with overlapping classes in a neural network is introduced. It allows one to estimate membership levels for every sample to every class. This network is the generalization of a vector neuro- and neuro-fuzzy Kohonen network and allows for data processing as they are fed in the on-line mode.

Using SOM in the performance monitoring of the emergency call-taking system

The collaborative emergency call-taking information system in the Czech Republic forms a network of cooperating emergency call centres processing emergency calls to the European 112 emergency number. Large amounts of various incident records are stored in its databases. The data can be used for mining spatial and temporal anomalies, as well as for the monitoring and analysis of the performance of the emergency call-taking system. In this paper we describe a method for knowledge discovery and visualisation targeted at the performance analysis of the system with respect to the organisation of the emergency call-taking information system and its data characteristics. The method is based on the Kohonen Self-Organising Map (SOM) algorithm and its extension, the Growing Grid algorithm.

A modified training scheme for SOFM to cluster multispectral images

In this paper, we propose modifications to Kohonen's Self-Organising Feature Map (SOFM) to achieve faster convergence specifically with respect to multispectral images. First, the raw image is pre-processed using data reduction technique to obtain reduced data set and then Condensed Nearest Neighbour (CNN) rule is applied to yield standard subset of samples. The samples in the standard subset are used to find the Best Matching Unit (BMU) and the samples in the reduced data set are used to update BMU and its neighbouring neurons. The SOFM is tested on: synthetic image data set and Harangi 1991, 1992 image data sets. Results are compared with conventional SOFM.

Reconfigurable Self-Organizing Neural Network Design and it's FPGA Implementation

The use of Kohonen self-organizing feature maps in real time applications requires high computational performance, especially for embedded systems and hence neural network chips are essential. A digital architecture of Kohonen neural network with learning capability and on-chip adaptation and storage is proposed with the implementation of Kohonen SelfOrganizing Map (SOM) neural networks on the low-cost Spartan-3 FPGAs. The architecture of this digital chip based on the idea that some assumptions for the restrictions of the algorithm can simplify the implementation. Using the Manhattan distance, a special treatment of the adaptation factor, and neighborhood functions will decrease the necessary chip area so that a high number of processing elements can be integrated on one chip.

A computational approach for the health care market

In this work we analyze the market for health care through a computational approach that relies on Kohonen's Self-Organizing Maps, and we observe the competition dynamics of health care providers versus those of patients. As a result, we offer a new tool addressing the issue of hospital behaviour and demand mechanism modelling, which conjugates a robust theoretical implementation together with an instrument of deep graphical impact.

Self-organizing fuzzy clustering neural network and application to electronic countermeasures effectiveness evaluation

A self-organizing fuzzy clustering neural network by combining the self-organizing Kohonen clustering network with the fuzzy theory is proposed. This network model is designed for the effectiveness evaluation of electronic countermeasures, which not only exerts the advantages of the fuzzy theory, but also has a good ability in machine learning and data analysis. The subjective value of sample versus class is computed by the fuzzy computing theory, and the classified results obtained by self-organizing learning of Kohonen neural network are represented on output layer. Meanwhile, the fuzzy competition learning algorithm keeps the similar information between samples and overcomes the disadvantages of neural network which has fewer samples. The simulation result indicates that the proposed algorithm is feasible and effective.

Improved surface roughness as a result of free-form surface machining using self-organized neural network

This paper is concerned with the free-form surface reorganization and assessment of a free-form model complexity, grouping particular surface geometrical properties within patch boundaries using self-organized Kohonen neural network (SOKN). Coordinate values of point cloud distributed at a particular surface were used as a surface properties descriptor, which was fed into SOKN where representative neurons for curvature, slope and spatial surface properties were established. On the basis of this approach, the surface patch boundaries were reorganized in such a manner that the finished machining strategies gave the best possible surface roughness results. The patch boundaries were constructed in accordance with the Gaussian and mean curvature, in order to achieve a smooth transition between patches, and in this way, preserve or even improve the desired curve and surface continuities (C 2 and G 2). It is shown that by reorganization of the boundaries in respect of curvature, slope and spatial point distribution, the surface quality of the finished free-form surface is improved. This approach was experimentally verified on 22 free-form models which were reorganized by SOKN and machined with finish milling tool-path strategies. The results show rather good improvement of the mean surface roughness profile R a for reorganized surfaces, when compared with unorganized surfaces.

Comprehensive ecological zoning of the Kuibyshev Reservoir

Multidimensional mathematical analysis of a set of abiotic and biotic indices was carried out on the basis of many-year (1958–1984, 1992) observations of the Kuibyshev Reservoir. The character of the space and time dynamics of the samples is studied; the localization of sections of statistically homogeneous data was implemented. Graphs of the long-term spatial taxonomic structure of water masses were analyzed for different periods. The graphs had been formed with the use of the principal component method, neuron network modeling, and self-organizing Kohonen maps. Comprehensive zoning of the reservoir water area was carried out with the help of GIS technologies.

Activation-Based Recursive Self-Organising Maps: A General Formulation and Empirical Results

We generalize a class of neural network models that extend the Kohonen Self-Organising Map (SOM) algorithm into the sequential and temporal domain using recurrent connections. Behaviour of the class of Activation-based Recursive Self-Organising Maps (ARSOM) is discussed with respect to the choice of transfer function and parameter settings. By comparing performances to existing benchmarks we demonstrate the robustness and systematicity of the ARSOM models, thus opening the door to practical applications.

Effective palette indexing for image compression using self-organization of Kohonen feature map

The process of limited-color image compression usually involves color quantization followed by palette re-indexing. Palette re-indexing could improve the compression of color-indexed images, but it is still complicated and consumes extra time. Making use of the topology-preserving property of self-organizing Kohonen feature map, we can generate a fairly good color index table to achieve both high image quality and high compression, without re-indexing. Promising experiment results will be presented.

Color reduction and estimation of the number of dominant colors by using a self-growing and self-organized neural gas

A new method for color reduction in a digital image is proposed, which is based on the development of a new neural network classifier and on a new method for Estimation of the Most Important Classes (EMIC). The proposed neural network combines the features of the well-known Growing Neural Gas (GNG) and the Kohonen Self-Organized Feature Map (KSOFM) neural networks. We call the new neural network Self-Growing and Self-Organized Neural Gas (SGONG). This combination produces a new neural network with outstanding features. The proposed technique utilizes the GNG mechanism of growing the neural lattice and the KSOFM leaning adaptation mechanism. Besides, introducing a number of criteria that have an effect on inserting or removing neurons, it is able to automatically define the number of the created neurons and their topology. Moreover, applying the EMIC method, the produced classes can be filtered and the most important classes can be found. The combination of SGONG and EMIC results in retaining the isolated and significant colors with the minimum number of color classes. The above techniques are able to be fed by both color and spatial features. For this reason a similarity function is used for vector comparison. The method is applicable to any type of color images and it can accommodate any type of color space.

Use of unsupervised and supervised artificial neural networks for the identification of lactic acid bacteria on the basis of SDS-PAGE patterns of whole cell proteins

Conventional multivariate statistical techniques (hierarchical cluster analysis, linear discriminant analysis) and unsupervised (Kohonen Self Organizing Map) and supervised (Bayesian network) artificial neural networks were compared for as tools for the classification and identification of 352 SDS-PAGE patterns of whole cell proteins of lactic acid bacteria belonging to 22 species of the genera Lactobacillus, Leuconostoc, Enterococcus, Lactococcus and Streptococcus including 47 reference strains. Electrophoretic data were pre-treated using the logistic weighting function described by Piraino et al. [Piraino, P., Ricciardi, A., Lanorte, M. T., Malkhazova, I., Parente, E., 2002. A new procedure for data reduction in electrophoretic fingerprints of whole-cell proteins. Biotechnol. Lett. 24, 1477-1482]. Hierarchical cluster analysis provided a satisfactory classification of the patterns but was unable to discriminate some species ( Leuconostoc, Lb. sakei/ Lb. curvatus, Lb. acidophilus/ Lb. helveticus, Lb. plantarum/ Lb. paraplantarum, Lc. lactis/ Lc. raffinolactis). A 7 × 7 Kohonen self-organizing map (KSOM), trained with the patterns of the reference strains, provided a satisfactory classification of the patterns and was able to discriminate more species than hierarchical cluster analysis. The map was used in predictive mode to identify unknown strains and provided results which in 85.5% of cases matched the classification obtained by hierarchical cluster analysis. Two supervised tools, linear discriminant analysis and a 23:5:2 Bayesian network were proven to be highly effective in the discrimination of SDS-PAGE patterns of Lc. lactis from those of other species. We conclude that data reduction by logistic weighting coupled to traditional multivariate statistical analysis or artificial neural networks provide an effective tool for the classification and identification of lactic acid bacteria on the basis of SDS-PAGE patterns of whole cell proteins.

APPROACH OF IMPLEMENTATION OF AN INTELLIGENT ENTERPRISE STAFF ADAPTIVE E-LEARNING

This paper presents the principles of development of an intelligent e-curricula model for the employees of the future generations of manufacturing or service processing systems. To introduce an adaptive e-learning property for the employees of an intelligent future enterprise, the proposed e-curricula model is constructed. It is based both on using an adaptive multi – agent system and on applying the following unsupervised learning algorithms: the Q-learning, and a special type of artificial neural network, the Kohonen's Self-Organizing Maps.

Better Kohonen Neural Network in Radar Images Compression

Maritime radar image may be a basis for a perspective ship position fixing system. Images obtained from navigational radars possess usually great amount of information. Using all of these information is usually practically impossible. It is necessary to compress information imbedded in each image to the size acceptable from the practical point of view. Such an effect may be obtained by application of feature extraction methods. This article presents one of such methods that is based on capabilities of modified self-organizing Kohonen neural network.

Synergistic Techniques for Better Understanding and Classifying the Environmental Structure of Landscapes

The desire to capture natural regions in the landscape has been a goal of geographic and environmental classification and ecological land classification (ELC) for decades. Since the increased adoption of data-centric, multivariate, computational methods, the search for natural regions has become the search for the best classification that optimally trades off classification complexity for class homogeneity. In this study, three techniques are investigated for their ability to find the best classification of the physical environments of the Mt. Lofty Ranges in South Australia: AutoClass-C (a Bayesian classifier), a Kohonen Self-Organising Map neural network, and a k-means classifier with homogeneity analysis. AutoClass-C is specifically designed to find the classification that optimally trades off classification complexity for class homogeneity. However, AutoClass analysis was not found to be assumption-free because it was very sensitive to the user-specified level of relative error of input data. The AutoClass results suggest that there may be no way of finding the best classification without making critical assumptions as to the level of class heterogeneity acceptable in the classification when using continuous environmental data. Therefore, rather than relying on adjusting abstract parameters to arrive at a classification of suitable complexity, it is better to quantify and visualize the data structure and the relationship between classification complexity and class homogeneity. Individually and when integrated, the Self-Organizing Map and k-means classification with homogeneity analysis techniques also used in this study facilitate this and provide information upon which the decision of the scale of classification can be made. It is argued that instead of searching for the elusive classification of natural regions in the landscape, it is much better to understand and visualize the environmental structure of the landscape and to use this knowledge to select the best ELC at the required scale of analysis.

The use of Markovian metapopulation models: Reducing the dimensionality of transition matrices by self-organizing Kohonen networks

Markovian population models are used in conservation biology to find an accurate estimate of a population's extinction probability. Such models require handling of large transition matrices and calculations are thus extremely time-consuming when large populations have to be studied. To accomplish these problems, some authors have suggested to group together several states/sizes of the population. Unfortunately, this so-called binning frequently results in errors in estimates obtained. The main problem with binning is that it assumes that grouped states behave nearly identical with respect to the underlying stochastic population process and that so far binning methods implicitly violate this assumption. In this paper, we present a new binning method based on self-organizing Kohonen neural networks for time-homogeneous Markovian metapopulation models. The neural networks are used to analyse one-step transitions of the Markov chain in order to group only nearly identical states. We show that the new method is more qualified for the use in conservation than deterministic methods that we had discussed in a previous paper (first order Fibonacci binning, pairs binning). It reveals more accurate and more reliable estimates than these methods. Errors in estimated extinction probabilities that were introduced by binning did not exceed one order of magnitude and errors in the global population size did not exceed 30%. These errors in estimates correspond to a low inaccuracy in model parameter values of population growth and migration ranging from ±1 to ±10%. The reduction in the state space of the studied metapopulations ranged from 21 to 33% per subpopulation. The resulting decrease in computing time caused by our binning method is substantial particularly with regard to simulations tasks such as comparing the extinction risk of several populations or performing a detailed sensitivity analysis for model parameters assumed. The successful estimation of the extinction risk of eight natural butterfly populations demonstrates the applicability of our new binning method in conservation biology. A comparison of extinction probabilities and mean population sizes estimated by Monte Carlo simulations [Griebeler, E.M., Seitz, A., 2002. An individual based model for the conservation of the endangered Large Blue butterfly, Maculinea arion (Lepidoptera: Lycaenidae). Ecol. Model. 134, 343–356] with those obtained from the respective Markov chains for these butterfly populations revealed similar results on the accuracy of estimates and the reduction in transition matrices that were predicted by the comparative error analysis for binning methods.

A neural networks approach to image data compression

We present a novel neural model for image compression called the direct classification (DC) model. The DC is a hybrid between a subset of the self-organizing Kohonen (SOK) model and the adaptive resonance theory (ART) model. The DC is a fast and efficient neural classification engine. The DC training utilizes the accuracy of the winner-takes-all feature of the SOK model and the elasticity/speed of the ART1 model. The DC engine has experimentally achieved much better results than the state-of-the-art peer image compression techniques (e.g., JPEG2000 and DjVu wavelet technology) especially in the domains of colored documents and still satellite images. We include a comprehensive analysis of the most important parameters of our DC system and their effects on system performance.