Self-Organizing Map

An Explorative Full Text An Explorative Full Text Information Retrieval Method Based Information Retrieval Method Based on SOM (Self Organizing Map) on SOM (Self Organizing Map) Algorithm to Order Documents Algorithm to Order Documents Based on Their Full Text Contents

Abstract—Satellite images often require segmentation in the presence of uncertainly which caused due to factors like environmental condition, poor resolution and poor illumination. Image processing applications depends on the quality of segmentation. This paper proposes a novel methodology namely “Satellite Image Segmentation using Energetic Self Organizing Map” (SIS-ESOM) method. This method can be used to improve the accuracy level of the satellite image segmentation. This segmentation method is also tolerable against noises in satellite images. This paper describes the implementation of two novel algorithms, namely Dynamic Adaptive Threshold based Background Optimization (DATBO) method and Energetic SOM (ESOM). The input image is undergone to fuzzy based noise Removal and DATBO image enhancement method. The optimum training samples of Energetic SOM are gathered by reduction of training vectors using Fuzzy C Means (FCM).The computed weight values of SOM are converted into transformed values using Discrete Wavelet Transform (DWT) and Discrete Cosine Transform (DCT). This DWT and DCT transformed weight values are hold much energy. The SOM testing process is applied in an energetic way using DWT & DCT transforms and a new square root(√𝟐) based similarity measurement method. The new SOM method is called Energetic SOM because it uses the energy transforms such as DWT and DCT. The segmented image obtained from this ESOM is further refined to get fine segmentation. Good segmentation performance can be possible in satellite images with higher PSNR.

(2) Where xi is the ith input vector, Wi,j is the weight vector connecting input i to output neuron j and Dj is the sum of Euclidian distance between input sample xi and it's connecting weight vector to jth output neuron which is called a map unit. There are different applications for SOM neural networks in WSNs routing protocols. These applications can be divided into three general groups: deciding optimal route, selection of cluster heads and clustering of nodes. The authors in [2] used Kohonen SOM neural networks for clustering and their analysis to study unpredictable behaviors of network parameters and applications.Clustering of sensor nodes using Kohonen Self Organizing Map (KSOM) is computed for various numbers of nodes by taking different parameters of sensor node such as direction, position, number of hops, energy levels, sensitivity, latency, etc. Cordina and Debono [6] proposed a new LEACH like routing protocol in which the election of Cluster Heads is done with SOM neural networks where SOM inputs are intended parameters for cluster heads. LEA2C apply the connectionist learning by the minimiza- tion of the distance between the input samples

Kohonen’s Self-Organizing Map uses an arranged set of neurons usually in 2-D rectangular or hexagonal grid [31]. Data reduction into 2-D dimensionality from high dimensionality is effective in approximation of similarity relations [32] [33] and also useful for data visualization to help in determining classes or similar patterns [34]. SOM transfers the arbitrary dimensions of incoming input data signals into one or two-dimensional map [35], and learn or discover the underlying structure of the input data. SOM has two layers of neurons; an input layer and an output layer. Each input vector is fully connected to each neuron in the output layer. The choice of the SOM grid size is determined by the degree of details of the findings; more generalization of finding requires less grid size than more detailed one [36].

This paper focused on the vision-based approach for recognizing static hand gestures representing letters of the alphabet, enter, space and backspace keys captured using a CMU camera and converted to computer readable form by image processing and used as input for the training, setting boundaries and testing of Kohonen Self-Organizing Map.

E-learning is the resulting product from the evolution of internet technology. It acts as a medium of learning virtually without limitation of time and space and the need for teachers to be present physically. Currently, Moodle which is a learning management system has become an important medium to deliver e-learning easily by providing customized tool for educators to deploy learning materials in various forms, provide online discussion forum, online quizzes, online assignments and online activities among students. Moodle capture the student’s interactions and activities while learning on-line using the log files. The data stored in the log files contain meaningful information such as the student’s behavior, preferences and knowledge level. This information is very useful for educators to analyze the student’s characteristics in order to improve the teaching methods. In addition, the student’s progress can be improved by overcome the problem of one-size-fits-all and also to improve student learning experienced while using the system. In this paper, the student’s action and behavior while using e-learning system are analyzed in order to identify the significant pattern of the student’s behavior using Self-Organizing Map (SOM) clustering technique. The ability of SOM to analyze large amounts of data with variety types of variables and with better visualization of the result give an advantage to this technique. The experiment shows that unsupervised learning using SOM is able to identify several clusters from the student’s behavior by visualization of high dimensional data into two-dimensional (2-D) space.

The implementation of SOM algorithms to deal with complicated data has at- tracted considerable attention from many researchers [3]-[9]. [10] [11] intro- duced concept of SOM, followed by [11] make development and applications. [12] [13] make applications to Robtics, [14] [15] give examples in Geo-Self-Organizing Map (GEO-SOM). In [16], constructive SOM are called SAM-SOM family. SOM can be used for exploring the clustering of genes in the medical field [17].

The self-organizing map (SOM) is a class of neural networks that are trained in an unsupervised manner using a competitive learning [12]. The neural gas is a biologically inspired adaptive algorithm [13]. The algorithm was named “neural gas” because of the dynamics of the vectors during the adaptation process which distribute themselves like a gas within the data space. A codebook M is an array of vectors. The dimensionality of the vectors is such as that of the analyzed vectors X l , l = 1,..., m , i.e., equal to n. The array M = { M M 1 , 2 ,..., M s } is one-dimensional in NG,

Abstract. In the paper, text mining and visualization by self-organizing map (SOM) are investigated. At first, textual information must be converted into numerical one. The results of text mining and visualization depend on the conversion. So, the influence of some control factors (the common word list and usage of the stemming algorithm) on text mining results, when a document dictionary is created, is investigated. A self-organizing map is used for text clustering and graphical representation (visualization). A comparative analysis is made where a dataset consists of scientific papers about the optimization, based on Pareto, simplex, and genetic algorithms. Two new measures are also proposed to estimate the SOM quality when the classified data are analyzed: distances between SOM cells, corresponding to data items assigned to the same class, and the distance between centers of SOM cells, corresponding to different classes. The quantization error is measured to estimate the SOM quality, too.

In this paper, an algorithm is proposed based on self- organizing map (SOM). This algorithm accepts the structure of a social network as its input in the form of a neighborhood graph of the social network. By applying changes in the learning phase of SOM network, by adjusting the weight of neurons of the network, divides the social network into different clusters. Based on the results of applying this algorithm on various social networks, it can be observed that this algorithm is effectively capable of clustering a social network.

Abstract: This paper presents an analysis for Self Organizing Map (SOM) using Response Surface Methodology (RSM) and Desirability Function to find the optimal parameters to improve performance. This comparative explores the relationship between explanatory variables (numerical and categorical) such a competitive algorithm and learning rate and response variables as training time and quality metrics for SOM. Response surface plots were used to determine the interaction effects of main factors and optimum conditions to the performance in classification of partial discharge (PD).

We have discussed the performance of the MRI image in terms of weight vector, execution time and tumor pixels detected. We have described several methods in medical image processing and discussed requirements and properties of techniques in brain tumor detection .our project is used to give more information about brain tumor detection and segmentation. The target area is segmented and the evaluation of this tool from the doctor, whom the project is cooperated with, is positive and this tool helps the doctors in diagnosis, the treatment plan making and state of the tumor monitoring. In future, the system should be improved by adapting more segmentation algorithm to suit the different medical image segmentation. Using color based image segmentation; it is possible to reduce the computational cost avoiding feature calculation for every pixel in the image. We have proposed a system for MR image of brain tumor, based on Self-organizing map and fuzzy C-mean Algorithm. Artificial vision aims to replace the human vision in various areas. Image analysis and interpretation represent an essential phase in the chain of the vision process by computer the design of the system depends on the Self-organizing map and fuzzy C mean algorithm to detect brain tumor, the efficiency have been proved by the results obtained from our project.

T he Kohonen Self-Organizing Map (KSOM) is one of the Neural Network unsupervised learning algorithms. This algorithm is used in solving problems in various areas, especially in clustering complex data sets. Despite its advantages, the KSOM algorithm has a few drawbacks; such as overlapped cluster and non-linear separable problems. Therefore, this paper proposes a modified KSOM that inspired from pheromone approach in Ant Colony Optimization. The modification is focusing on the distance calculation amongst objects. The proposed algorithm has been tested on four real categorical data that are obtained from UCI machine learning repository; Iris, Seeds, Glass and Wisconsin Breast Cancer Database. From the results, it shows that the modified KSOM has produced accurate clustering result and all clusters can clearly be identified .

Applying occurrence filter with 9×9 window to 10 individual landform element maps and using the resulting texture maps as input to a SOM allowed the mapping of 4 major landforms including mountains with steep slopes, plane areas with gentle slopes, dissected ridges and lower valleys with moderate to very steep slopes and main valleys with gentle to moderate slopes. The results show that Self Organizing Map is a very promising and efficient tool for geomorphological studies. There was a very good agreement between identified landforms and contour lines. This new procedure is encouraging and offers new possibilities to study both types of terrain features, general landforms and landform elements.

In order to amend the above-mentioned deficiency, L´ opez-Rubio et al. [8] proposed a Principal Components Analysis Self-Organizing Map (PCASOM), where the manifold learning is realized by using an incremental PCA. However computations related to the updating of covariance matrices and the correspond- ing eigenproblem make PCASOM computationally expensive. Liu [9] devised an Adaptive Manifold Self-Organizing Map (AMSOM) as an extension of the basic algorithm of the ASSOM, which attempts to learn linear manifolds. The AM- SOM was applied to face recognition and demonstrated superior performance to the standard PCA method as shown in [9]. An extension of AMSOM that uses a kernel method to account for nonlinear manifolds was proposed in [10].

The work is related to the use of Self Organizing Map (SOM) which is a type of unsupervised Artificial Neural Network (ANN), as an aid to Maximal Ratio Combining (MRC) in order to improve bit error rate (BER) values of demodulated signals in wireless channels that have both Gaussian and multipath fading characteristics. Among the architectures and algorithms suggested for ANN, the SOM has the special property of effectively creating spatially organised “ internal representations” of various features of input signals and their abstractions. The advantage of using the SOM is that it doesn’t require any reference signal for training. Modulation technique used in this work is Bipolar Phase Shift Keying (BPSK) in Gaussian and multipath Rayleigh fading channels. The work adopts ANN block as part of a MRC set-up and is tested under SNR variation between -10 to 10 dB in Gaussian and multipath fading channels. The results generated justify the use of SOM neural network block as an aid to the MRC setup.

Distribution transformer is the transformer that steps down the voltage from primary voltage of the electric distribution to the customer. Here, the transformer is provide alternating current power by electronic magnetic induction using a primary to secondary winding where it in the same frequency [3]. From this transformer the FRA and DGA method will used to measure the condition of the transformer. The data from both measurements will apply to Self Organizing map where this software can read the data, explore it and make a decision for the data given.

Self-Organizing Map (SOM) is one of the widely ap- plied neural networks and has some interesting features over other neural networks. One advantages of using SOM is that it is quite robust with respect to noisy data, and its advantages over other classification models are its natural robustness and its very good illustrative power. Indeed, it has been successfully applied for several clas- sification tasks.

the performance of the K-Means algorithm. A major limitation with fuzzy ART is the unrestricted growth of clusters. Fuzzy ART networks sometimes produce numerous clusters each with only a small number of members. The authors have used fuzzy ART as an initial seed generator and K-Means as the final clustering algorithm. In [15], Santosh K et al. have developed a clustering algorithm that groups users according to their Web access patterns. The algorithm is based on the ART1 version of adaptive resonance theory. ART1 offers an unsupervised clustering approach that adapts to the changes in user’s access patterns over time without losing earlier information. It applies specifically to binary vectors. They have compared their algorithm’s performance with the traditional K-Means clustering algorithm and showed that the ART1-based technique performed better in terms of intra cluster distances. In [16], Antonia S et al. have developed a variant of the classical SOM called Growing Hierarchical SOM (GHSOM). They have suggested a new visualization technique for the patterns in the hierarchical structure. In [17, 18], T. Kohonen et al. have used Kohenen’s Self Organizing Map to organize Web documents into a two dimensional map according their document content. Documents which are similar in content are located in similar regions on the map. In [19], Kate A. Smith et al. have developed LOGSOM, a system that utilizes Kohonen’s Self Organizing Map to organize Web page in to two dimensional maps. The organization of the Web pages is based on the user’s navigation behavior rather than the content of the Web pages.

As for consuming oil, the participants were categorized into 2 groups: 1) people who mostly consumed saturated oil (hydrogenated vegetable fats, butter or other animal fats, margarine) and 2) those who mainly consumed unsaturated oil (vegetable oil or frying oil). The following questions were used to evaluate the status of smoking, and alcohol and tobacco use: “Do you smoke any kind of cigarette at this time? (Any type of factory-made cigarettes, hand-made cigarettes, or cigars)”, “Do you use the hookah at this time?” and “Did you drink alcoholic beverages in the past year?” Statistical analysis: The mean and standard error were reported for quantitative variables, and frequency and percentage were used for qualitative variables. About 3.32% of the cells had missing data. Missing data were imputed performing single imputation and regression model in the MICE package in the R software (13-14). The self- organizing map (SOM) was applied for clustering (15-18). SOM is a nonparametric and single-layer artificial neural network clustering technique to group similar individuals (the details of this method are reported in appendix A). The silhouette index was used to decide the optimum number of clusters (19).