Efficient Self-organizing Map Research Articles

Stereo Imaging Using Hardwired Self-Organizing Object Segmentation.

Stereo vision utilizes two cameras to acquire two respective images, and then determines the depth map by calculating the disparity between two images. In general, object segmentation and stereo matching are some of the important technologies that are often used in establishing stereo vision systems. In this study, we implement a highly efficient self-organizing map (SOM) neural network hardware accelerator as unsupervised color segmentation for real-time stereo imaging. The stereo imaging system is established by pipelined, hierarchical architecture, which includes an SOM neural network module, a connected component labeling module, and a sum-of-absolute-difference-based stereo matching module. The experiment is conducted on a hardware resources-constrained embedded system. The performance of stereo imaging system is able to achieve 13.8 frames per second of 640 × 480 resolution color images.

An efficient SOM and EM-based intravascular ultrasound blood vessel image segmentation approach

Intravascular ultrasound (IVUS) is a catheter-based imaging method used in the study of atherosclerotic disease. IVUS produces cross-sectional images of the blood vessels that enable quantitative assessment of the plaque. Automatic segmentation of the anatomical structures in the IVUS image is a really challenging task due to the presence of noise and catheter artifacts. Hence, this paper presents an efficient self-organizing map (SOM) and expectation-maximization (EM)-based approach for the segmentation of cross-sectional view of the IVUS blood vessel image. In our proposed work, the directional filtering is used to improve the signal to noise ratio of the blood vessel image. The Hough transform is used for predicting the circle in the image. Segmentation of the image is performed using the SOM and EM algorithm. After the segmentation process, extraction of the common pixels is performed. Gray-level co-occurrence matrix is applied for extracting features from the image. Fuzzy-relevance vector machine based classification of the image is performed. From the comparison results, it is clearly observed that the proposed approach is highly efficient than the existing techniques.

An efficient self-organizing map (E-SOM) learning algorithm using group of neurons

AbstractIn the learning process of the conventional SOM, the neuron which is closer to the winner neuron learns more than the neuron which is farther away from the winner neuron. The neurons farther away from input are not able to learn properly and some dead units are left on the map. To decrease dead unit problem and improve the learning efficiency, an efficient Self-organzing map algorithm using group of neurons has been proposed. In this paper, we have divided the neurons on the map into two groups according to distance from input: normal and distant. The neurons which are far away from the input have been named distant neurons. We have done some changes in the kernel function for the distant neurons and then compared the learning efficiency of the algorithms by applying on standard input dataset. The results have been compared using three well known parameters, which are widely accepted for checking the learning efficiency of machine learning algorithms. It has been observed from the experimental res...

Special Issue on Innovation over Multimedia Processing

With broadband networks, audio/video coding, and processing techniques and user devices swiftly advancing, multimedia streaming over networks is now a reality. Distributed interactive multimedia applications ? one of the fastest growing market sectors demanding innovations ? are covered in this special issue on multimedia processing (MP). This special issue focuses on ambitious and intriguing papers from experts on a wide variety of multimedia areas. These articles address innovations and offer effective solutions to MP problems. Y. Fushio et al. propose a shadow generation system based on shape recognition, while M. Suzuki et al. develop musical expression generation reflecting the userfs impression in kansei space and fuzzy rules. K. Ishizuka et al. evaluate an operetta songs generation system based on the impressions given by story scenes, which are strongly related to the integrative research projects of kansei engineering, music, text, and visual perception. T. Toyota et al. develop a web news visualization system based on an efficient self-organizing map using a restricted region search and dimensionality reduction. S. Kawauchi et al. propose a knowledge expansion support system using related search keyword generation based on Wikipedia categories and point-wise mutual information, involving that are aggressive looking progressive research projects covering huge text information. T. Tanabata et al. present an interactive data mining tool for microarray data analysis using formal concept analysis, which adds a new dimension to the bio-informatics field. T. Kawamura et al. analyze keywords input to a book search engine. T. Sugimoto et al. propose a recommendation system involving the use of comprehensive trend indications based on weighted complete graphs, which are related to topical discussions in the information retrieval and recommendation field. Readers are sure to find new and valuable information and ideas in their own fields in these excellent articles.

Visualization of the Internet News Based on Efficient Self-Organizing Map Using Restricted Region Search and Dimensionality Reduction

In this paper, we propose a system to visualize the relationships in huge quantities of Internet news by twodimensional self-organizing maps instead of the conventional methods of listing Internet news. In the proposed method, morphological analysis is conducted on the texts of Internet news to generate input vectors with elements of keywords. The characteristics specific to Internet news that many of the vector elements become sparse allows dimensional reductions as well as speeding up of self-organizing mapping with restricted search regions in learning. We verify through evaluation experiments with the data of 80 pieces of news that the proposed system can reduce computation time by 75% to 99% and can create more efficient SOM compared with the generally available SOM.

An Efficient Initialization Scheme for SOM Algorithm Based on Reference Point and Filters

This paper proposes an efficient self-organizing map algorithm based on reference point and filters. A strategy called Reference Point SOM (RPSOM) is proposed to improve SOM execution time by means of filtering with two thresholds T1 and T2. We use one threshold, T1, to define the search boundary parameter used to search for the Best-Matching Unit (BMU) with respect to input vectors. The other threshold, T2, is used as the search boundary within which the BMU finds its neighbors. The proposed algorithm reduces the time complexity from O(n2) to O(n) in finding the initial neurons as compared to the algorithm proposed by Su et al.[16]. The RPSOM dramatically reduces the time complexity, especially in the computation of large data set. From the experimental results, we find that it is better to construct a good initial map and then to use the unsupervised learning to make small subsequent adjustments.

An efficient self-organizing map designed by genetic algorithms for the traveling salesman problem

As a typical combinatorial optimization problem, the traveling salesman problem (TSP) has attracted extensive research interest. In this paper, we develop a self-organizing map (SOM) with a novel learning rule. It is called the integrated SOM (ISOM) since its learning rule integrates the three learning mechanisms in the SOM literature. Within a single learning step, the excited neuron is first dragged toward the input city, then pushed to the convex hull of the TSP, and finally drawn toward the middle point of its two neighboring neurons. A genetic algorithm is successfully specified to determine the elaborate coordination among the three learning mechanisms as well as the suitable parameter setting. The evolved ISOM (eISOM) is examined on three sets of TSP to demonstrate its power and efficiency. The computation complexity of the eISOM is quadratic, which is comparable to other SOM-like neural networks. Moreover, the eISOM can generate more accurate solutions than several typical approaches for TSP including the SOM developed by Budinich, the expanding SOM, the convex elastic net, and the FLEXMAP algorithm. Though its solution accuracy is not yet comparable to some sophisticated heuristics, the eISOM is one of the most accurate neural networks for the TSP.