In general cases of pattern recognition, a pattern to be recognized is first represented by a set of features and the measured values of the features are then classified. Finding features relevant to recognition is thus an important issue in recognizer design. As a fundamental design framework taht systematically enables one to realize such useful features, the Subspace Method (SM) has been extensively used in various recognition tasks. However, this promising methodological framework is still inadequate. The discriminative power of early versions was not very high. The training behavior of a recent discriminative version called the Learning Subspace Method has not been fully clarified due to its empirical definition, though its discriminative power has been improved. To alleviate this insufficiency, we propose in this paper a new discriminative SM algorithm based on the Minimum Classification Error/Generalized Probabilistic Descent method and show that the proposed algorithm achieves an optimal accurate recognition result, i.e., the (at least locally) minimum recognition error situation, in the probabilistic descent sense.

Methods of automatically adjusting delay time and feedback gain in controlling chaos by delayed feedback control are proposed. These methods are based on a gradient-descent procedure minimizing the squared error between the current state and the delayed state. The method of adjusting delay time and that of adjusting feedback gain are applied to controlling chaos in numerical calculations of Rossler Equation and Duffing equation, respectively. Both methods are confirmed to be successful.

It is well known that offset errors in the multipliers of neural LSIs can have fatal effects on performance. The aim of this study is to understand theoretically how offset errors affect performance of neural LSIs. We have used a single-layer perceptron as an example, and compare our theoretically derived results with computer simulations. We have found that offset errors in the multipliers for the forward process can be canceled out through learning, but those for the updating process cannot be. We have examined the asymptotic behavior of learning for the updating process and derived a mathematical expression for dL, the excess of the averaged loss function L. The derived expression gives us a basis for estimating robustness with respect to the offset errors. Our analysis indicates that dL can be expressed in the form of a quadratic form of offset errors and the inverse of the Hessian matrix of L. We have found that increasing the number of synapses degrades the performacne. We have also learned that enlarging the input signal level and reducing the signal level of the desired response can be effective techniques for reducing the effects of offset errors of the updating process.

In order to construct fuzzy systems automatically, there are many studies on combining fuzzy inference with neural networks. In these studies, fuzzy models using self-organization and vector quantization have been proposed. It is well known that these models construct fuzzy inference rules effectively representing distribution of input data, and not affected by increment of input dimensions. In this paper, we propose a destructive fuzzy modeling using neural gas network and demonstrate the validity of a proposed method by performing some numerical examples.

Detailed investigations of the microstructural properties of SrGa2S4:Ce3+ thin films grown by deposition from binary vapors (DBV) were carried out by X-ray diffraction analysis (XRD), energy dispersive X-ray diffraction measurements (EDX), electron probe microanalysis (EPMA), and X-ray photoelectron spectroscopy (XPS) depth profiling. The results indicate uniform distribution of the constituent elements in the nearly stoichiometric structure of the thin films. Photoluminescence (PL) data including absorption and luminescence spectra in the temperature range of 10 to 300 K and decay characteristics show that an increase in Ce concentration from 0.2 to 3 mol% is accompanied with a marked increase in both the intensity of activator absorption and decay time, while the emission and excitation bands remain fixed in position. A mechanism involving the concentration-dependent interactions between different centers in the lattice is proposed, which may explain the experimentally observed behavior.

Electroluminescent (EL) devices with mixed single layer that consist of fluorescent dyes, distylylbiphenyl derivative (DPVBi) and triphenylamine derivative (TPD), are studied. Blue light emission was observed from the device with DPVBi and TPD. White emission over 2,500 cd/m2 was observed from the devices with mixed single layer of DPVBi, TPD and dicyanomethylene derivative (DCM).

There is now a world-wide trend towards the downsizing of information systems using a number techniques, such as clientserver architecture. Consequently, enterprise networks are fast growing in terms of size and functionality. These networks need to be managed effectively. Researchers have been working on the development of management solutions for enterprise networks, using recent advances in software engineering, communication protocols, and artificial intelligence techniques. However, not much work has been published on the role of human factors in the integrated management of networks and systems. This paper presents a new Cooperative management Methodology for Enterprise Networks (CoMEN), based on Computer Supported Cooperative Work (CSCW) techniques.

Neural networks can be used as associative memories which can learn problems of acquiring input-output relations presented by examples. The learning time problem addresses how long it takes for a neural network to learn a given problem by a learning algorithm. As a solvable model to this problem we analyze the learning dynamics of the linear associative memoty with the least-mean-square algorithm. Our result shows that the learning time τ of the linear associative memory diverges in τ (1-ρ)-2 as the memory rate ρ approaches 1. It also shows that the learning time exhibits the exponential dependence on ρ when ρ is small.

This paper reviews analog LSI design issues for optical transmission applications; covering ultra-high-speed transmission over 10 Gb/s, multi-Gb/s systems, optical interconnection systems, and optical access. In the future system development, further advancements in not only optical device technology but also LSI technology are eagerly required. More and more sophisticated circuit design techniques are needed to lower power and operation voltage, increase integration, eliminate external elements and adjustments.

RF integration, until recently the integration of active devices in conventional architectures suitable for discrete-component circuits, is now turning into full-integration based on new architectures developed specifically for an LSI technology. This paper reviews some of the key existing and emerging circuit techniques and discusses the serious problem of crosstalk. In order to develop miniature and low power RF transceivers, direct-conversion and monolithic VCO's will be further studied. Silicon bipolar technology will still be playing major role beyond the year 2,000, and CMOS will also be used in certain applications.

We demonstrate highly quick response and long distance distributed oil sensors using a newly developed eccentric core fiber (ECF). This distributed oil sensor,based on an interaction between measurand oil and evanescent-wave from the ECF, has achieved as short as 4 minutes response time by using an improved coating material and a sensing length over 17 km at a signal wavelength of 1310 nm. The observed sensitivity characteristics coincide with the calculations of the evanescent power outside the cladding and it is shown that the sensitivity can be well estimated from the amount of the outer cladding component of the evanescent power.

In this letter, a theoretical estimation of pick-up characteristics of the fiber probe of Photon Scanning Tunneling Microscopy based on the Wiener-Hopf technique taken account of the weakly guiding approximation are reported. As a result, it is found that diffracted waves by the extremity of the fiber probe mainly act on the mode excitation rather than transmitted waves, then the pick-up characteristics are well accordance with typical experiments quality and quantity.

In-line transceiver chip emitting at 1.3 µm and receiving at 1.55 µm is described regarding the crosstalk occurring between the 1.3-µm laser and the 1.55-µm integrated photodiode. Contribution of optical and electrical crosstalk to the overall crosstalk is measured and discussed. Techniques to overcome the crosstalk are proposed and demonstrated by showing the feasibility of system compatible diplex sensitivities.

In this paper we discuss the limiting behavior of the search direction of the steepest descent method in minimizing the Rayleigh quotient. This minimization problem is equivalent to finding the smallest eigenvalue of a matrix. It is shown that the search direction asymptotically alternates between two directions represented by linear combinations of two eigenvectors of the matrix. This is similar to the phenomenon in minimizing the quadratic form. We also show that these eigenvectors correspond to the largest and second-smallest eigenvalues, unlike in the case of the quadratic form.

To realize a low-cost WDM transceiver module based on a PLC-platform, simple, assembly techniques have been successfully developed. The formation of index marks with an accuracy of below 0.1 µm has made it possible to mount Opto-electronic devices on the silicon terrace of the PLC-platform by a passive alignment. A newly developed trench formation technique for inserting a 1.3/1.5 µm WDM dielectric filter enabled us not only to ensure a stable WDM function but also to prevent excess loss associated with the dielectric filter scheme. It is found that these two technologies are practically useful to achieve high-performance WDM transceiver module.

A simple extension to treat evanescent modal excitation at the aperture of a parallel plane waveguide is shown here by GTD diffracted rays with complex propagation angles. Numerical comparison with other solution confirmed that our simple solution can be used for modal excitation estimation below the cut-off frequency.

Because 60 GHz frequency band has been allotted for the research and development purpose of millimeter wave systems in Japan, various circuit components and systems have been fabricated by using printed transmission lines. The NRD guide (nonradiative dielectric waveguide) is another candidate as a transmission medium for millimeter wave integrated circuit applications since its performance has been shown to be excellent in this frequency band. This paper is concerned with the development of a 60 GHz digital transceiver for millimeter wave LAN use based on NRD guide technologies. The trans-ceiver consists of frequency stabilized Gunn oscillator, circulator, PIN diode modulator, balanced mixer, directional coupler and transmitting and receiving pyramidal horn antennas. The notable advantages of the circuit components are the high reliability of the Gunn oscillator, the wide bandwidth of the circulator, and the high frequency operation of the PIN diode modulator beyond 100 Mbps. Interference between transmitted and received signals, which must be caused by coupling between transmitting and receiving antennas, is eliminated by simple techniques such as introducing filters in the base band and IF circuits. By using NRD guide digital transceivers, both-way data transmission between two computers can be achieved simultaneously and a 60 GHz wireless LAN system has been developed successfully.

This paper discusses our newly developed technology for making GaAs/InGaAs/GaAs Tetrahedral-Shaped Recess (TSR) quantum dots. The heterostructures were grown by low-pressure MOVPE in tetrahedral-shaped recesses created on a (111) B oriented GaAs substrate using anisotropic chemical etching. We examined these structures by using cathodoluminescence (CL) measurements, and observed lower energy emissions from the bottoms of, and higher energy emissions from the walls of the TSRs. This suggests carrier confinement at the bottoms with the lowest potential energy. We carried out microanlaysis of the structures by using TEM and EDX, and found an In-rich region that had grown vertically from the bottom of the TSR with a (111)B-like bond configuration. We also measured a smaller diamagnetic shift of the lower energy photoluminecscence (PL) peak in the structure. Based on these results, we have concluded that the quantum dots are formed at the bottoms of TSRs, mainly because of the dependence of InAs composition on the local crystalline structure in this system. We also studied the lateral distribution and vertical alignment of TSR quantum dots by CL and PL measurements respectively. The advantages of TSR quantum dot technology can be summarized as follows: (i) better control in dot positioning in the lateral direction, (ii) realization of dot sizes exceeding limitations posed by lithography, (iii) high uniformity of dot size, and (iv) vertical alignment of quantum dots.

The electromagnetic force of evanescent field acting on dielectric slab is studied with the use of Maxwell stress tensor. The results show that dielectrics slab may receive always an attractive force when the incident wave is evanescent field while a pressure or an attractive force when the wave is propagating one. The magnitude of the attractive force by evanescent field is much larger than that of the propagating wave. And here some numerical examples are given.

Video services such as video-on-demand are expected to be a motivation for deploying multimedia services in residential areas. These services should increase customer demand for video channels as customer demands become more sophisticated and diverse in the future. Therefore, it is important to determine how network configurations (i.e. network transition scenarios) should evolve in response to changes in access network demand. This paper proposes economical deployment of access networks based on transition scenarios. We conclude that transition scenarios offer more economical deployment than single-network configurations. Two transition scenarios, from passive double-star to fiber single-star, and from hybrid fiber-coax to fiber single-star, are evaluated as examples. These transition scenarios are economical even when customer demand changes. The transition starting time affects the present worth of annual charges (PWAC) of access networks more than the transition period does.