Xiaoqiu WANG Hua LIN Jianming LU Hiroo SEKIYA Takashi YAHAGI
This paper presents a compensating method based on Self-Organizing Map (SOM) for nonlinear distortion, which is caused by high-power amplifier (HPA) in 16-QAM-OFDM system. OFDM signals are sensitive to nonlinear distortions and different methods are studied to solve them. In the proposed scheme, the correction is done at the receiver by a SOM algorithm. Simulations are carried out considering an additive white Gaussian (AWG) transmission channel. Simulation results show that the SOM algorithm brings perceptible gains in a complete 16-QAM-OFDM system.
Takeshi TATEYAMA Seiichi KAWATA Hideaki OHTA
In this paper, a new grouping method for Group Technology using Self-Organizing Map (SOM) is proposed. The purpose of our study is to divide machines in a factory into any number of cells so that the machines in each cell can process a similar set of parts to increase productivity. A main feature of our method is to specify not only the number of the cells but also the maximum and minimum numbers of machines in a cell. Some experimental results show effectiveness of our proposed algorithm.
Shinya FUKUMOTO Noritaka SHIGEI Michiharu MAEDA Hiromi MIYAJIMA
Neural networks for Vector Quantization (VQ) such as K-means, Neural-Gas (NG) network and Kohonen's Self-Organizing Map (SOM) have been proposed. K-means, which is a "hard-max" approach, converges very fast. The method, however, devotes itself to local search, and it easily falls into local minima. On the other hand, the NG and SOM methods, which are "soft-max" approaches, are good at the global search ability. Though NG and SOM exhibit better performance in coming close to the optimum than that of K-means, the methods converge slower than K-means. In order to the disadvantages that exist when K-means, NG and SOM are used individually, this paper proposes hybrid methods such as NG-K, SOM-K and SOM-NG. NG-K performs NG adaptation during short period of time early in the learning process, and then the method performs K-means adaptation in the rest of the process. SOM-K and SOM-NG are similar as NG-K. From numerical simulations including an image compression problem, NG-K and SOM-K exhibit better performance than other methods.
Self-organizing map is a widely used tool in high-dimensional data visualization. However, despite its benefits of plotting very high-dimensional data on a low-dimensional grid, browsing and understanding the meaning of a trained map turn to be a difficult task -- specially when number of nodes or the size of data increases. Though there are some well-known techniques to visualize SOMs, they mainly deals with cluster boundaries and they fail to consider raw information available in original data in browsing SOMs. In this paper, we propose our Factor controlled Hierarchical SOM that enables us select number of data to train and label a particular map based on a pre-defined factor and provides consistent hierarchical SOM browsing.
This paper describes a method of analyzing musical sound using a self-organizing map. To take compound factors into account, energy spectra whose frequency ranges were based on the psycho-acoustic experiments were used as input data. Results for music compact discs confirmed that our method could effectively display the positioning and relationships among musical sounds on a map.
Masaaki KUBO Zaher AGHBARI Kun Seok OH Akifumi MAKINOUCHI
This paper proposes a technique for indexing, clustering and retrieving images based on their edge features. In this technique, images are decomposed into several frequency bands using the Haar wavelet transform. From the one-level decomposition sub-bands an edge image is formed. Next, the higher order auto-correlation function is applied on the edge image to extract the edge features. These higher order autocorrelation features are normalized to generate a compact feature vector, which is invariant to shift, image size. We used direction cosine as measure of distance not to be influenced by difference of each image's luminance. Then, these feature vectors are clustered by a self-organizing map (SOM) based on their edge feature similarity. The performed experiments show higher precision and recall of this technique than traditional ways in clustering and retrieving images in a large image database environment.
This paper describes an analysis of the electromagnetic interference (EMI) aspects of electrostatic discharge (ESD), which sometimes causes serious damage to electrical systems. To classify EMI-related properties resulting from ESD events, we used a self-organizing neural network, which can map high-dimensional data into simple geometric relationships on a low-dimensional display. Also, to clarify the effect of a high-speed moving discharge, we generated one-shot discharges repeatedly and measured the ESD current in the time domain to obtain its EMI-related characteristics of this phenomenon. Based on the measured data, we examined several differential properties of ESD waveforms including the maximum amplitude and energy level, and analyzed these multi-dimensional data using the self-organizing neural network scheme. The results showed that the high-speed moving discharges can increase the maximum amplitude, relative energy, and entropy of ESD events, and that the positioning of the EMI level of each ESD event can be effectively visualized in a two-dimensional space.
Newaz M. S. RAHIM Takashi YAHAGI
Finite-state vector quantization (FSVQ) is a well-known block encoding technique for digital image compression at low bit rate application. In this paper, an improved feature map finite-state vector quantization (IFMFSVQ) algorithm using three-sided side-match prediction is proposed for image coding. The new three-sided side-match improves the prediction quality of input blocks. Precoded blocks are used to alleviate the error propagation of side-match. An edge threshold is used to classify the blocks into nonedge or edge blocks to improve bit rate performance. Furthermore, an adaptive method is also obtained. Experimental results reveal that the new IFMFSVQ reduces bit rate significantly maintaining the same subjective quality, as compared to the basic FMFSVQ method.
Xiaoqiu WANG Hua LIN Jianming LU Takashi YAHAGI
Recently, neural networks (NNs) have been extensively applied to many signal processing problem due to their robust abilities to form complex decision regions. In particular, neural networks add flexibility to the design of equalizers for digital communication systems. Recurrent neural network (RNN) is a kind of neural network with one or more feedback loops, whereas self-organizing map (SOM) is characterized by the formation of a topographic map of the input patterns in which the spatial locations (i.e., coordinates) of the neurons in the lattice are indicative of intrinsic statistical features contained in the input patterns. In this paper, we propose a novel receiver structure by combining adaptive RNN equalizer with a SOM detector under serious ISI and nonlinear distortion in QAM system. According to the theoretical analysis and computer simulation results, the performance of the proposed scheme is shown to be quite effective in channel equalization under nonlinear distortion.
Michiharu MAEDA Hiromi MIYAJIMA
In this paper we treat a novel adaptation strength according to neighborhood ranking of self-organizing neural networks with the objective of avoiding the initial dependency of reference vectors, which is related to the strength in the neural-gas network suggested by Martinetz et al. The present approach exhibits the effectiveness in the average distortion compared to the conventional technique through numerical experiments. Furthermore the present approach is applied to image data and the validity in employing as an image coding system is examined.
This paper describes a QoS evaluation method for VoIP communications using a self-organizing neural network. Based on measurements in real environments, evaluation results confirmed that our method can effectively display total QoS level composed of several QoS-related factors such as PSQM+ and end-to-end delay.
Hua LIN Xiaoqiu WANG Jianming LU Takashi YAHAGI
A signal suffers from nonlinear, linear, and additive distortion when transmitted through a channel. Linear equalizers are commonly used in receivers to compensate for linear channel distortion. As an alternative, novel equalizer structures utilizing neural computation have been developed for compensating for nonlinear channel distortion. In this paper, we propose a neural detector based on self-organizing map (SOM) in a 16 QAM system. The proposed scheme uses the SOM algorithm and symbol-by-symbol detector to form a neural detector, and it adapts well to the changing channel conditions, including nonlinear distortions because of the topology-preserving property of the SOM algorithm. According to the theoretical analysis and computer simulation results, the proposed scheme is shown to have better performance than traditional linear equalizer when facing with nonlinear distortion.
Xiaoqiu WANG Hua LIN Jianming LU Takashi YAHAGI
Detection of nonlinearly distorted signals is an essential problem in telecommunications. Recently, neural network combined conventional equalizer has been used to improve the performance especially in compensating for nonlinear distortions. In this paper, the self-organizing map (SOM) combined with the conventional symbol-by-symbol detector is used as an adaptive detector after the output of the decision feedback equalizer (DFE), which updates the decision levels to follow up the nonlinear distortions. In the proposed scheme, we use the box distance to define the neighborhood of the winning neuron of the SOM algorithm. The error performance has been investigated in both 16 QAM and 64 QAM systems with nonlinear distortions. Simulation results have shown that the system performance is remarkably improved by using SOM detector compared with the conventional DFE scheme.
Tomoyuki HIGUCHI Genshiro KITAGAWA
A hierarchical structure of the statistical models involving the parametric, state space, generalized state space, and self-organizing state space models is explained. It is shown that by considering higher level modeling, it is possible to develop models quite freely and then to extract essential information from data which has been difficult to obtain due to the use of restricted models. It is also shown that by rising the level of the model, the model selection procedure which has been realized with human expertise can be performed automatically and thus the automatic processing of huge time series data becomes realistic. In other words, the hierarchical statistical modeling facilitates both automatic processing of massive time series data and a new method for knowledge discovery.
We analyze the dynamics of self-organizing cortical maps under the influence of external stimuli. We show that if the map is a contraction, then the system has a unique equilibrium which is globally asymptotically stable; consequently the system acts as a stable encoder of external input stimuli. The system converges to a fixed point representing the steady-state of the neural activity which has as an upper bound the superposition of the spatial integrals of the weight function between neighboring neurons and the stimulus autocorrelation function. The proposed theory also includes nontrivial interesting solutions.
Takeshi YAMAKAWA Keiichi HORIO
In this letter, the novel mapping network named self-organizing relationship (SOR) network, which can approximate the desired I/O relationship by employing the modified Kohonen's learning law, is proposed. In the modified Kohonen's learning law, the weight vectors are updated to be attracted to or repulsed from the input vector.
Taira NAKAJIMA Hiroyuki TAKIZAWA Hiroaki KOBAYASHI Tadao NAKAMURA
We propose a learning algorithm for self-organizing neural networks to form a topology preserving map from an input manifold whose topology may dynamically change. Experimental results show that the network using the proposed algorithm can rapidly adjust itself to represent the topology of nonstationary input distributions.
Michiharu MAEDA Hiromi MIYAJIMA Sadayuki MURASHIMA
This paper describes an adaptive neural vector quantization algorithm with a deleting approach of weight (reference) vectors. We call the algorithm an adaptive learning and self-deleting algorithm. At the beginning, we introduce an improved topological neighborhood and an adaptive vector quantization algorithm with little depending on initial values of weight vectors. Then we present the adaptive learning and self-deleting algorithm. The algorithm is represented as the following descriptions: At first, many weight vectors are prepared, and the algorithm is processed with Kohonen's self-organizing feature map. Next, weight vectors are deleted sequentially to the fixed number of them, and the algorithm processed with competitive learning. At the end, we discuss algorithms with neighborhood relations compared with the proposed one. The proposed algorithm is also good in the case of a poor initialization of weight vectors. Experimental results are given to show the effectiveness of the proposed algorithm.
We discuss possible new principles of information processing by utilizing microscopic, semi-microscopic and macroscopic phenomena occuring in nature. We first discuss quantum mechanical universal information processing in microscopic world governed by quantum mechanics, and then we discuss superconducting phenomena in a mesoscopic system, especially an information processing system using flux quantum. Finally, we discuss macroscopic self-organizing phenomena in biology and suggest possibility of self-organizing devices.
Young Pyo JUN Hyunsoo YOON Jung Wan CHO
The self-organizing feature map is one of the most widely used neural network paradigm based on unsupervised competitive learning. However, the learning algorithm introduced by Kohonen is very slow when the size of the map is large. The slowness is caused by the search for large map in each training steps of the learning. In this paper, a fast learning algorithm based on incremental ordering is proposed. The new learning starts with only a few units evenly distributed on a large topological feature map, and gradually increases the number of units until it covers the entire map. In middle phases of the learning, some units are well-ordered and others are not, while all units are weekly-ordered in Kohonen learning. The ordered units, during the learning, help to accelerate the search speed of the algorithm and accelerate the movements of the remaining unordered units to their topological locations. It is shown by theoretical analysis as well as experimental analysis that the proposed learning algorithm reduces the training time from O(M2) to O(log M) for M by M map without any additional working space, while preserving the ordering properties of the Kohonen learning algorithm.