Yen-Tseng Hsu

Grey self-organizing feature maps

In each training iteration of the self-organizing feature maps (SOFM), the adjustable output nodes can be determined by the neighborhood size ofthe winning node. However, it seems that the SOFM ignores some important information, which is the relationships that actually exist between the input training data and each adjustable output node, in the learning rule. By viewing input data and each adjustable node as a reference sequence and a comparative sequence, respectively, the grey relations between these sequences can be seen. This paper thus incorporates the grey relational coe8cient into the learning rule ofthe SOFM, and a grey clustering method, namely the GSOFM, is proposed. From the simulation results, we can see that the best result ofthe proposed method applied f or analysis ofthe iris data outperf orms those ofother known unsupervised neural network models. Furthermore, the proposed method can e:ectively solve the traveling salesman problem. c 2002 Elsevier Science B.V. All rights reserved.

Forecasting the turning time of stock market based on Markov-Fourier grey model

This paper presents an integration prediction method including grey model (GM), Fourier series, and Markov state transition, known as Markov-Fourier grey model (MFGM), to predict the turning time of Taiwan weighted stock index (TAIEX) for increasing the forecasting accuracy. There are two parts of forecast. The first one is to build an optimal grey model from a series of data, the other uses the Fourier series to refine the residuals produced by the mentioned model. Finally, the Markov state scheme is used for predicting the possibility of location results to promote the intermediate results generated by the Fourier grey model (FGM). It is evident that the proposed approach gets the better result performance than that of the other methods.

High-precision forecast using grey models

In recent years the grey theorem has been successfully used in many prediction applications. The proposed Markov-Fourier grey model prediction approach uses a grey model to predict roughly the next datum from a set of most recent data. Then, a Fourier series is used to fit the residual error produced by the grey model. With the Fourier series obtained, the error produced by the grey model in the next step can be estimated. Such a Fourier residual correction approach can have a good performance. However, this approach only uses the most recent data without considering those previous data. In this paper, we further propose to adopt the Markov forecasting method to act as a longterm residual correction scheme. By combining the short-term predicted value by a Fourier series and a long-term estimated error by the Markov forecasting method, our approach can predict the future more accurately. Three time series are used in our demonstration. They are a smooth functional curve, a curve for the stock market and th...

A novel image compression using grey models on a dynamic window

Based on the grey model (GM), a simple and fast methodology is developed for lossy image compression. First of all, the image is decomposed into some different-size image windows through the judgement of grey difference level; then the GM (1,1) of grey system theory is used as a fitter to model those window pixels. The proposed algorithms can be contrasted with the conventional compression techniques such as discrete cosine transform or vector quantization (VQ) algorithms in their dynamic modelling sequence and flexible block size. Especially, the compression and decompression process do not require an extra decoder and only utilize the modelling parameters to reconstruct the image by reversing the operation of GM (1,1). Experiments with some (512 x 512) images indicate that not only the average bit number per pixel and peak signal-to-noise ratio but also the coding time and decoding time of this lossy image compression algorithm based on GM (1,1) are better than those of block truncation coding with VQ.

A GreyART system for grey information processing

Abstract In this paper, grey relational theory is successfully fused into the ART 1 model to construct GreyART models such as GreyARTarea and GreyARTslope. Both models possess the structure and learning ability of an ART-type NN while using grey relational analysis to process the grey information among the patterns to cluster. GreyART models could be placed in the family of ART-type NNs, the input information of which is extended from binary patterns to multi-valued patterns. Additionally, GreyART models accede to the property of ART-type NN in which the stored prototype is self-organized as a reference template during the real-time learning process. Such a property of GreyART model solves the problems in which the reference patterns have to be pre-determined in most of the grey classification methods. Computer simulations were conducted using a given pattern to examine the geometric grey-characteristic of GreyART-classifying.

Grey-neural forecasting system

In this paper, a new nonlinear forecasting system using a grey predictor model and neural network tuner is proposed. This paper puts its emphasis on a few data and incomplete information to build the predictive system, excavates the connotative essence of a signal from the GM (1,1) predictor and refers to anomalistic (over predictive error) conditions to build the neural tuner database. So in the forecasting model the GM (1,1) model system predictive value will be appreciably modified while the anomalistic condition occurs. Simulation in well-known Mackey-Glass time series is presented to demonstrate the performance of the proposed predictive system.

Profit refiner of futures trading using clustering algorithm

Lowering psychological pressure of investors and increasing the futures trading profit are the main purposes of this paper. First of all, this study aims to transfer profit curve (PC) generated by a non-AI-based trading strategy into technical indices, and enable clustering of high-low points of PC to display high-low point signals through some AI-based methods such as Grey Clustering, SOM and K-mean. Next, it attempts to close the transaction with high-point signal, and then open a position at low-point one, thus constructing three groups of profit refiners: GCR (Grey Clustering Refiner), SOMR (SOM Refiner) and KMR (K-Mean Refiner). Finally, the features of these refiners are analyzed to evaluate the test results using some performance indices. SOMR could improve the profit to the greatest possible extent, followed by GCR and KMR; on the other hand, KMR could lower psychological pressure, followed by SOMR and GCR. As a whole, three groups of refiners can really improve the profit and alleviate psychological burden of investors.

A G2LA vector quantization for image data coding

In this paper, based on the Grey theory, a novel measurement method in a large volume and high dimension of information system is proposed for vector quantization (VQ) design and applied to image data coding. In the VQ coding procedure, it is often needs several epochs of clustering and always fails to obtain a better codebook; for instance, the well-known generalized Lloyd algorithm (GLA) easily traps into suboptimal codebook and does not have the ability to locate an optimal codebook during any clustering iteration with a random initial codebook. Hence, we propose a G^2LA design to solve heavy times of clustering procedure and at least to gain the best suboptimal codebook. In order to avoid edge degradation, firstly, the new selection of initial codevectors is adopted as the fast grey vector quantization (FGVQ) procedure which chooses nonhomogeneous vectors from a large volume image data. Then extending the GLA to G^2LA method by utilizing the measurement of grey relational analysis (GRA) which depends on the effect of relative objective and initial codevectors of FGVQ to obtain a better representative codebook. Experiment results show that at the same bit rate the G^2LA has not only the quickly convergence time but also high quality reconstructed image than traditional GLA technique with Euclidean distance measure, especially in high dimension and a large volume data system.