Chun-g Lin

Effective Learning Rate Adjustment of Blind Source Separation Based on an Improved Particle Swarm Optimizer

Blind source separation (BSS) is a technique used to recover a set of source signals without prior information on the transformation matrix or the probability distributions of the source signals. In previous works on BSS, the choice of the learning rate would result in a competition between stability and speed of convergence. In this paper, a particle swarm optimization (PSO)-based learning rate adjustment method is proposed for BSS, and a simple decision-making method is introduced for how the learning rate should be applied in the current time slot. In the experiments, samples of four and ten source signals were mixed and separated and the results were compared with other related approaches. The proposed approach exhibits rapid convergence, and produces more efficient and more stable independent component analysis algorithms, than other related approaches.

A hierarchical classification system for sleep stage scoring via forehead EEG signals

The study adopts the structure of hierarchical classification to develop an automatic sleep stage classification system using forehead (Fpl and Fp2) EEG signals. The hierarchical classification consists of a preliminary wake detection rule, a novel feature extraction method based on American Academy of Sleep Medicine (AASM) scoring manual, feature selection methods and SVM. After estimating the preliminary sleep stages, two adaptive adjustment schemes are applied to adjust the preliminary sleep stages and provide the final estimation of sleep stages. Clinical testing reveals that the proposed automatic sleep stage classification system is about 77% accuracy and 67% kappa for individual 10 normal subjects. This system could provide the possibility of long term sleep monitoring at home and provide a preliminary result of sleep stages so that doctor could decide if a patient needs to have a detailed diagnosis using Polysomnography (PSG) system in a sleep laboratory of hospital.

PSO-based learning rate adjustment for blind source separation

Blind source separation (BSS) is a technique for recovering a set of source signals without a priori information on the transformation matrix or the probability distributions of the source signals. In the previous works of BSS, the choice of the learning rate would reflect a trade-off between the stability and the speed of convergence. In this paper, we adapted the particle swarm optimization (PSO) technique to find suitable learning rates for each signal in each time slot. Experiments employing four mixed source signals were separated by our work and compared with other related approaches. The proposed approach exhibited rapid convergence and made the independent component analysis (ICA) algorithms become more efficient and stable than other related approaches.

Cluster Distance Factor Searching by Particle Swarm Optimization for Self-Growing Radial Basis Function Neural Network

A very important step for the RBF network training is to decide a proper number of hidden nodes. This paper proposes a PSO-based algorithm for searching optimal cluster distance factor. Thus, the self-growing RBF network training algorithm can be realized by employing the optimal cluster distance factor to create hidden neurons automatically from the input data set. Two experiments that include function approximation and chaotic time series prediction are used to compare our proposed method with other related approaches. The PSO-based RBF network exhibits better generalization performance and shorter training time.

A Radial Basis Function Neural Network with Adaptive Structure via Particle Swarm Optimization

Radial Basis Function neural network (RBFNN) is a combination of learning vector quantizer LVQ-I and gradient descent. RBFNN is first proposed by (Broomhead & Lowe, 1988), and their interpolation and generalization properties are thoroughly investigated in (Lowe, 1989), (Freeman & Saad, 1995). Since the mid-1980s, RBFNN has been used to apply on many applications, such as pattern classification, system identification, nonlinear function approximation, adaptive control, speech recognition, and time-series prediction, and so on. In contrast to the well-known Multilayer Perceptron (MLP) Networks, the RBF network utilizes a radial construction mechanism. MLP were trained by the error Back Propagation (BP) algorithm, since the RBFNN has a faster training procedure substantially and adopts typical two-stage training scheme, it can avoid solution to fall into local optima. A key point of RBFNN is to decide a proper number of hidden nodes. If the hidden node number of RBFNN is too small, the generated output vectors may be in low accuracy. On the contrary, it with too large number of hidden nodes may cause over-fitting for the input data, and influences global generalization performance. In conventional RBF training approach, the number of hidden node is usually decided according to the statistic properties of input data, then determine the centers and spread width for each hidden nodes by means of k-means clustering algorithm (Moddy & Darken, 1989). The drawback of this approach is that the network performance is depended on the pre-selected number of hidden nodes. If an unsuitable number is chosen, RBFNN may present a poor global generalization capability, as slow training speed, and requirement for large memory space. To solve this problem, the self-growing RBF techniques were proposed in (Karayiannis & Mi, 1997), (Zheng et al, 1999). However, the predefined parameters and local searching on solution space cause the inaccuracy of approximation from a sub-solution. Evolutionary computation is a globally optimization technique, where the aim is to improve the ability of individual to survive. Among that, Genetic Algorithm (GA) is a parallel searching technique that mimics natural genetics and the evolutionary process. In (Back et al, 1997), they employed GA to determine the RBFNN structure so the optimal number and distribution of RBF hidden nodes can be obtained automatically. A common approach is applied GA to search for the optimal network structure among several candidates

Cluster-based Adaptive Mutation Mechanism To Improve the Performance of Genetic Algorithm

This paper discusses the improvement of premature convergence in genetic algorithm (GA) used for optimizing multimodal numerical problems. Mutation is the principle operation in GA for enhancing the degree of population diversity, but is not efficient often, particularly in traditional GA. Moreover, the definition of mutation rate is a tradeoff between computing time and accuracy. In our work, we introduce the cluster method nearest neighborhood for estimating population diversity. According to this estimation, the mutation rate is adaptively given and repeat chromosomes are discarded over evolution. Consequently, the proposed cluster-based GA can choose a suitable mutation number for reducing computing time and maintain the population variety for preventing premature convergence. It is confirmed in numerical optimization simulations that the proposed GA is superior than traditional GA used fixed mutation rate in terms of accuracy, computing time and convergent speed

A self-organized neural network for blind separation process with unobservable sources

A self-organized rule is proposed to process the separation of unknown number of sources of blind signal. The rule is applied to the feed-forward neural network (FFNN) which is based on a mean weighting gradient (MWG) function. The algorithms adjust the architecture of neural network. For both, the performances are the mixture of node number and MWG threshold /spl xi/.The separating number of sources estimated accurately by experiment of computer simulations.

Hybrid multi-model forecasting system: a case study on display market

This paper provides a novel hybrid multi-model forecasting system, with a special focus on the changing regional market demand in the display markets. Through an intensive case study of the ups and downs of the display industry, this paper examines the panel makers suffered from low panel price and unstable market demand, then they have changed to react to the rapid demand in the market or have lower panel stock for keeping supply and demand more balanced. In addition, this paper suggests a co-evolution forecasting process of sales and market factor. It can automatically apply various combinations of both linear and nonlinear models, and which alternatives deliver the lowest statistical error and produce a good estimate for the prediction of markets. Moreover, this article shows how the system is modeled and its accuracy is proved by means of experimental results; and judged by 3 evaluation criteria, including the mean square error (MSE), the mean absolute percentage error (MAPE), and the average square root error (ASRE) were used as the performance criteria to automatically select the optimal forecasting model. Finally, the results showed that the proposed system had considerably better predictive performance than previous and individual models. To summarize, the proposed system can reduce the users effort for easier obtaining the desired forecasting results and create high quality forecasts.

Fuzzy Neural Network-Based Influenza Diagnostic System

As certain diseases are characterized by subjective perceptions of described symptoms, if symptoms are not obvious, physicians can easily mistake them for other illnesses. In order to assist physicians to quickly and accurately diagnose results, a medical diagnostic aid expert system was put forth in this study. The system uses the fuzzy system, back-propagation neural network (BPNN), and fuzzy neural network (FNN) as the core engines of the influenza diagnostic expert system. The three systems were compared whereas the expert systems inferred output served as the data for the prognosis of occurrences of illnesses, thereby providing physicians a diagnostic reference and reducing diagnostic error rates in order to ensure early detections and treatment by doctors and prevent more serious illnesses that may arise due to complications.

A Rule-based Forecasting System Integrating combining and Single Forecast for Decision Making

Many studies demonstrated that combining forecasts produces consistent but modest gains in accuracy. However, little researches define well the conditions under neither which combining is most effective nor how methods should be combined in each situation. In this paper, a rule-based forecasting system industry is proposed in order to compare forecast performance between combining forecasts and single forecasts, then define these conditions and to specify more effective combinations, finally suggest the best methods. Two comparative case studies for the telecommunications and TFT-LCD industry are proposed to examine the performance of the proposed system. Results from this study indicate that combining forecasts outperform single forecasts only when data set is data have various nonlinear characteristics. In this research, empirical evidence shows that rules based on causal forces improved the selection of forecasting methods, the structuring of time series, and the assessment of prediction intervals.