Feature subset selection basically depends on the design of a criterion function to measure the effectiveness of a particular feature or a feature subset and the selection of a search strategy to find out the best feature subset. Lots of techniques have been developed so far which are mainly categorized into classifier independent filter approaches and classifier dependant wrapper approaches. Wrapper approaches produce good results but are computationally unattractive specially when nonlinear neural classifiers with complex learning algorithms are used. The present work proposes a hybrid two step approach for finding out the best feature subset from a large feature set in which a fuzzy set theoretic measure for assessing the goodness of a feature is used in conjunction with a multilayer perceptron (MLP) or fractal neural network (FNN) classifier to take advantage of both the approaches. Though the process does not guarantee absolute optimality, the selected feature subset produces near optimal results for practical purposes. The process is less time consuming and computationally light compared to any neural network classifier based sequential feature subset selection technique. The proposed algorithm has been simulated with two different data sets to justify its effectiveness.

In this paper, we are concerned in obtaining multicast trees in packet-switched networks such as ATM nets, when there exist constraints on the packet (cell)-replication capabilities of the individual switching nodes. This problem can be formulated as the Steiner tree problem with degree bounds on the nodes, so we call it the Degree-Constrained Steiner Tree problem (DCST). Four heuristic algorithms are proposed: the first is a combined version of two well-known Steiner tree algorithms, heuristic Naive and the shortest path heuristic (SPH), and the second is a relaxation algorithm based on a mathematical formulation of the DCST, and the last two use a tree reconfiguration scheme based on the concept of 'logical link. ' We experimentally compare our algorithms with the previous ones in three respects; number of solved instances, objective value or tree cost, and computation time. The experimental results show that there are few instances unsolved by our algorithms, and the objective values are mostly within 5% of optimal. Computation times are also acceptable.

In this paper, we consider the problem of active learning, and give a necessary and sufficient condition of sample points for the optimal generalization capability. By utilizing the properties of pseudo orthogonal bases, we clarify the mechanism of achieving the optimal generalization capability. We also show that the condition does not only provide the optimal generalization capability but also reduces the computational complexity and memory required to calculate learning result functions. Based on the optimality condition, we give design methods of optimal sample points for trigonometric polynomial models. Finally, the effectiveness of the proposed active learning method is demonstrated through computer simulations.

In this paper, we present a numerical method with guaranteed accuracy to solve initial value problems (IVPs) of normal form simultaneous first order ordinary differential equations (ODEs) which have wide domain. Our method is based on the algorithm proposed by Kashiwagi, by which we can obtain inclusions of exact values at several discrete points of the solution curve of ODEs. The method can be regarded as an extension of the Lohner's method. But the algorithm is not efficient for equations which have wide domain, because the error bounds become too wide from a practical point of view. Our purpose is to produce tight bounds even for such equations. We realize it by combining Kashiwagi's algorithm with the mean value form. We also consider the wrapping effects to obtain tighter bounds.

This paper describes the implementation of a proposed image filter into a Discrete-Time Cellular Neural Network (DT-CNN). The three stages that compose the filter are described, showing that the resultant filter is capable of (1) erasing or detecting several concentric shapes simultaneously, (2) thresholding and (3) thinning of gray-scale images. Because the DT-CNN has to fill certain conditions for this filter to be implemented, it becomes a modified version of a DT-CNN. Those conditions are described and also experimental results are clearly shown.

In this paper we show that the neuron filter is effective for relaxing the coefficient sensitiveness of the Hopfield neural network for combinatorial optimization problems. Since the parameters in motion equation have a significant influence on the performance of the neural network, many studies have been carried out to support determining the value of the parameters. However, not a few researchers have determined the value of the parameters experimentally yet. We show that the use of the neuron filter is effective for the parameter tuning, particularly for determining their values experimentally through simulations.

We often see reflection phenomenon in our life. For example, through window glass, we can see real objects, but reflection causes virtual objects to appear in front of the glass. Thus, it is sometimes difficult to recognize the real objects. Some works have been proposed to separate these real and virtual objects using an optical property called polarization. However, they have a restriction on one assumption: the angle of incidence. In this paper, we overcome this difficulty using independent component analysis (ICA). We show the efficiency of the proposed method, by experimental results.

This paper proposes a hardware configuration using only single pole dual throw (SPDT) switches to realize the previously proposed automatic calibration method using transmitting signals (ACT) for the adaptive array in TDD communication systems. The proposed configuration obtains the same calibration values as the conventional ACT does while reducing the number of switch branches. The transmission pattern using the proposed calibration method is also presented based on an experimental adaptive array testbed in an actual microcell environment. The experimental results show that the ideal radiation pattern formation is achieved by employing the proposed calibration method in an environment with a moving terminal station and where arriving co-channel interference exists.

Web caching has become an important problem when addressing the performance issues in Web applications. The expiration time of the Web data item is useful a piece of information for performance enhancement in Web caching. In this paper, we introduce the notion of the effective reference probability that incorporates the effect of expiration time for Web caching. For a formal approach, we propose the continuous independent reference model extending the existing independent reference model. Based on this model, we define formally the effective reference probability and derive it theoretically. By simply replacing the reference probability in the existing cache replacement algorithms with the effective reference probability, we can take the effect of expiration time into account. The results of performance experiments show that the replacement algorithms using the effective reference probability always outperform existing ones. In particular, when the cache fraction is 0.05 and data update is comparatively frequent (i.e., the update frequency is more than 1/10 of the reference frequency), the performance is enhanced by more than 30% in LRU-2 and 13% in Aggarwal's method. The results show that the effective reference probability significantly enhances the performance of Web caching when the expiration time is given.

In this paper, we propose new approaches to speech enhancement based on soft decision. In order to enhance the statistical reliability in estimating speech activity, we introduce the concept of a global speech absence probability (GSAP). First, we compute the conventional speech absence probability (SAP) and then modify it according to the newly proposed GSAP. The modification is made in such a way that the SAP has the same value of GSAP in the case of speech absence while it is maintained to its original value when the speech is present. Moreover, for improving the performance of the SAP's at voice tails (transition periods from speech to silence), we revise the SAP's using a hang-over scheme based on the hidden Markov model (HMM). In addition, we suggest a robust noise update algorithm in which the noise power is estimated not only in the periods of speech absence but also during speech activity based on soft decision. Also, for improving the SAP determination and noise update routines, we present a new signal to noise ratio (SNR) concept which is called the predicted SNR in this paper. Moreover, we demonstrate that the discrete cosine transform (DCT) enhances the accuracy of the SAP estimation. A number of tests show that the proposed method which is called the speech enhancement based on soft decision (SESD) algorithm yields better performance than the conventional approaches.

In this paper, a simple chaotic circuit using two RC phase shift oscillators and a diode is proposed and analyzed. By using a simpler model of the original circuit, the mechanism of generating chaos is explained and the exact solutions are derived. The exact expression of the Poincare map and its Jacobian matrix make it possible to confirm the generation of chaos using the Lyapunov exponents and to investigate the related bifurcation phenomena.

This paper describes a cumulative distribution function (CDF) of edge direction for detecting road lanes. Based on the assumptions that there are no abrupt changes in the direction and location of road lanes and that the intensity of lane boundaries differs from that of the background, the CDF is formulated, which accumulates the edge magnitude for edge directions. The CDF has distinctive peak points at the vicinity of lane directions due to the directional and the positional continuities of a lane. To obtain lane-related information, we construct a scatter diagram by collecting edge pixels, of which the direction corresponds to the peak point of the CDF, then perform the principal axis-based line fitting for the scatter diagram. Because noises can cause many similar features appear or disappear in an image, to prevent false alarms or miss detection, a recursive estimator of the CDF was introduced, and also a scene understanding index (SUI) was formulated by the statistical parameters of the CDF. The proposed algorithm has been implemented in real time on video data obtained from a test vehicle driven on a typical highway.

An analytic equation was derived for the time jitter of digital NRZ signals due to inter-symbol interference in the PCB transmission lines loaded by DRAM chips which are located in uniform spacing. The inter-symbol interference is caused by a low-pass filtering effect of the loaded transmission line. Good agreements were observed between the equation and measurements with an average error of 17.5%.

In this paper, we shall construct mathematical theory based on the concept of set-valued mappings, suitable for available operation of network systems extraordinarily complicated and diversified on large scales. Fundamental conditions for availability of system behaviors of such network systems are clarified in a form of fixed point theorem for system of set-valued mappings.

Skew angle effects on the transition noise are analyzed in the longitudinal disk media by micromagnetic simulations at area densities from 14.3 Gb/in2 to 31.5 Gb/in2. The transition noise, including the peak, width and jitter noise, is the dominant noise in ultra-high density disk recording systems. An isotropic medium and an oriented medium, with a fixed grain size of 135 and a coercivity of 2900 Oe, are chosen for the noise analysis. The peak noise is studied by the distribution of the peak magnetization amplitude Mp in each bit. The transition a-parameter is no longer the value as given in the William-Comstock approximation. It is found that the transition noise is highly dependent on both the linear den sity and the skew angle, where the bit length and the grain size are on the same order. In both media, the medium noise increases severely when the skew angle is above 10 degrees.

The analysis of the radiating properties of a multilayer structure where chirality is introduced is here addressed. Both the effects on the resonant behaviour and on the radiation patterns have been considered for different multilayer structures. The adopted procedure is full wave and leads to the numerical analysis performed via the Methods of Moment in the spectral domain.

Particulate media composed of very small particles were studied to determine high-density recording performance and thermal stability. Studied media included metal particulate media with mean particle length of 71, 102 and 148 nm, and Ba ferrite particulate media with mean diameter of 22, 28 and 50 nm. Using a loss-term simulation program, taking into account gap-loss, spacing-loss and particle length loss, the recording capability (D20 of 265 kFRPI for MP and 290 kFRPI for Ba ferrite media) was estimated. Thermal stability was evaluated from magnetization time decay measurements. It was found that MP media with large Ku values and 71 nm particles were satisfactorily stable, and the particle volume is still large enough in respect of thermal stability. However, 22-nm Ba ferrite media were less stable, primarily because of small Ku values and particle volume. It was also clarified that positive inter-particle interaction accelerates magnetization time decay, in the presence of a large reverse field.

Feasibility of a color shutter using ferroelectric liquid crystal polymer panel and a field sequential ultra high-resolution CRT with the color shutter as a color field-switching device was studied. The color shutter consists of ferroelectric liquid crystal polymer panels and color polarizers. First, evaluation indices of the color shutter, such as the color gamut, the average transmittance and the white chromaticity shift, were formulated, and the simulation of evaluation indices was examined, where the spectral transmittance characteristics of the polarizer were changed in steps. It was indicated that there was a tradeoff between the color gamut and the average transmittance of the color shutter, and the shutter configuration that provides 0.096 (63% to NTSC) color gamut and 4.3% average transmittance was selected based on the simulation results. Next, the three-line simultaneous scanning method of the monochrome CRT was improved so that the disturbance due to the raster modulation was eliminated by averaging the distribution of beam luminance apparently. To confirm results of the study, the prototype of 21-inch screen size was produced, and the following display characteristics was obtained: luminance of 71 cd/m2, contrast ratio of 146:1 and color gamut of 0.096 (63% to NTSC) under the standard room lighting environment.

A nonlinear Volterra-series analysis of multiple ion-implanted GaAs FETs is given that relates carrier profile parameters of ion-implantation to nonlinear rf characteristics of a FET. Expressions for nonlinear coefficients of transconductance are derived from drain current-voltage characteristics of a multiple ion-implanted FET. Nonlinear transfer functions (NLTFs) are then obtained using Volterra series approach. Using these NLTFs third-order intermodulation distortion and power gain are explicitly given. A good agreement has been found between the calculation and the measurement for a medium power GaAs FET with a total gate width of 800 µm operated at 10-dB back off, verifying the usefulness of the present analysis.

This paper shows a comparison between several procedures to represent the Physical Optics (PO) current density into a hybrid Moment-Method-Physical-Optics (MM-PO) code. Some numerical results demonstrate that a set of basis functions suitable for the Method of Moments (MM) may be inappropriate to model the PO currents. A new evaluation of the PO operator is proposed. The radiation can be analytically determined and, since it includes a linear interpolation of the phase, it can be applied over large triangular domains. This allows a drastic reduction of the computational cost, maintaining or even improving the level of accuracy.