The design theory was revealed by theoretical analysis of the measuring apparatus, and was confirmed experimentally. Higher quality tags having new circuit disigns were proposed by the revealed theory. The measuring apparatus equivalent to the security system was produced to estimate the properties of the LC resonant circuit security tags quantitatively.

Customers' requirements for telecommunications services have been changing from simple point-to-point and single-medium communications to multiple-point and multiple-media, as well as changes in bandwidth requirements. In addition, customers' end stations (ES) will not be homogeneous, from POTS-like terminal to sophisticated multimedia workstation that handles various kind of media. This paper discusses communications management architectures that provide multimedia multipoint communications in a heterogeneous ESs environment. It summarizes existing communications management architectures, and discusses advanced architectures that provide multimedia communications, multipoint communications and heterogeneity handling capabilities in a flexible and efficient manner. The objectives of this architecture are to integrate multimedia multipoint communications capabilities into the network, and provide customers with versatile communications service request interface which can handle from plain single-medium point-to-point communications to multimedia multipoint communications. This paper especially focuses on the management functionalities for multimedia multipoint service fabrics (MMSF) which may be owned by the network provider, the third-party service provider, or the customer. It discusses variations of MMSF management protocol implementations in the case that the MMSF is owned by either of the three. It also discusses implementation of customers' various operations requests to the MMSF in the communication phase.

In general, message passing multiprocessors suffer from communication overhead and shared memory multiprocessors suffer from memory contention. Also, data I/O overhead limits performance. In particular, computer vision tasks that require massive computation are strongly affected by these disadvantages. This paper proposes new parallel architectures for computer vision, a Flexibly (Tightly/Loosely) Coupled Multiprocessor (FCM) and a Flexibly Coupled Hypercube Multiprocessor (FCHM) to alleviate these problems. FCM and FCHM have a variable address space memory in which a set of neighboring memory modules can be merged into a shared memory by a dynamically partitionable topology. FCM and FCHM are based on two different topologies: reconfigurable bus and hypercube. The proposed architectures are quantitatively analyzed using computational models and parallel vision algorithms are simulated on FCM and FCHM using the Intel's Personal SuperComputer (iPSC), a hypercube multiprocessor, showing significant performance improvements over that of iPSC.

We are developing GENESIS, a new seamless total environment for designing, developing, installing, and operating various types of telecommunication networks as extremely large distributed processing applications in the future network integrated by ATM. Similar uniform architectures for quick introduction and easy management of service or operation applications have been proposed, such as by TINA, but there has been insufficient study on how to operate and con figure those applications. This paper discusses the implementation model and execution environment in GENESIS from the viewpoint of flexible operation according to network conditions. The implementation model can describe detailed configurations under various conditions on design or operation, independently of the execution environment. To achieve the goals of GENESIS, our execution environment provides message handling functions and a transparent interface for controlling network resources independently of the configuration, and dynamic reconfiguration functions that are independent of the execution. This paper also reports the prototype system GENESIS-1. The GENESIS-1 message handling mechanism and the effect of the reconfiguration functions are described.

In this paper, we propose the architecture of BALANCE (Better Adaptive Load-balancing through Acquiring kNowledge of Characteristic of an Environment) in which users can submit their jobs without acquiring either a status of an environment or characteristics of jobs and servers even in a widely connected heterogeneous network. The architecture of BALANCE includes three types of information bases and two types of daemons. Information bases, namely job, resource, and environment information base, manage the knowledge of job characteristics, available resources for CPUs, and status of the environment, respectively, as a proxy for users. The dispatching daemon selects an adequate server for each job using knowledge stored in the information bases. A service daemon executes each job. On completing each job, a service daemon gets a statistic of the job and returns it to the dispatching daemon where the job came from so that the statistic will be available at the next dispatching time. BALANCE enables an environment (1) to balance the load, (2) to share software functions as well as hardware facilities, and (3) to learn a user's job characteristics. We have implemented a prototype with more than 50 heterogeneous UNIX workstations connected by different networks. Two simple experiments on this prototype are presented. These experiments show a load balancing scheme that takes the characteristics of each job into account.

This paper proposes new real–time heuristic distributed parallel algorithms for search, which are based on the concepts of propagations and competitions of concurrent waves. These algorithms are characterized by simplicity and clearness of control strategies for search, and distinguished abilities in many aspects, such as real–time performance, wide suitability for searching AND/OR implicit graphs, and ease in hardware implementation.

With advances in the speed, bandwidth and reliability of telecommunications networks and in the performance of workstations, distributed processing has become widespread. Information sharing among distributed nodes and its mutual exclusion are of great importance for efficient distributed processing. This paper systematizes and quantitizes a shared memory called Data-Cyclic Shared Memory (DCSM) from the viewpoints of memory organization and access mode. In DCSM, the propagation delay of transmission lines and the data relaying delay in each node are used for information storage, and memory information encapsuled in the form of "memory cells" circulates infinitely in a logical ring type network. The distinctive feature of DCSM, in addition to the way data is stored, is that data and the access control are completely distributed, which contrasts with existing memory where both are centralized. Therefore, there are no performance bottlenecks caused by concentrating memory access. Distributed Shared Memory (DSM), which has a scheme similar to DCSM's, has also been proposed for distributed environments. In DSM, the data is also distributed but the control for accessing each data is centralized. From the viewpoints of memory organization and the access method, DCSM is very flexible. For instance, word length can be spatially varied by defining data size at each address, and each node can be equipped with mechanisms for special functions such as the content address specification and asynchronous report of change in contents. Because of this flexibility, it can be called a "software-defined memory." The analysis also reveals that DCSM has the disadvantages of large access delay and small memory capacity. The capacity can be enlarged by inserting FIFO type queues into the circulation network, and the delay can be shortened by circulating replicas of original memory cells. However, there is a trade off between the maximal capacity and the mean access time. DCSM has many potential applications, such as in the mutual exclusion control of distributed resources.

This paper describes that a neural network, which consists of neurons with piecewise–linear sigmoid characteristics, is able to approximate any piecewise–linear map with origin symmetry. The neural network can generate "deterministic diffusion" originating from its diffusive trajectory.

Numerical simulations for the (3+1)-dimensional optical-field dynamics of nonstationary pulsed beams that propagate down Kerr-like nonlinear channel waveguides are carried out for what is to our knowledge the first time. Time-resolved intrapulse switching due to spontaneous symmetry breaking of optical fields from a quasilinear symmetric field to a nonlinear asymmetric field is analyzed. A novel instability phenomenon triggered by the symmetry breakdown is found.

In an image fiber containing a large number of cores, a certain class of crosstalk has been found to decrease with the distance along the fiber axis. This crosstalk is absolutely distinguished from the usual crosstalk that increases with the distance. A theoretical model is presented based on the power transfer between three groups of modes supported by each core. The process of power transfer is described by coupled power equations. Values of the coupling coefficients can be determined from the measurement of the crosstalk. The equations are solved numerically for the transmission of a point image. The results are in good agreement with measurement results.

We have classified parenchymal echo patterns of cirrhotic liver into 3 types, according to the size of hypoechoic nodular lesions. We have been studying an ultrasonic image diagnosis system using the three–layer back–propagation neural network. In this paper, we will describe the applications of the neural network techniques for recognizing and classifying chronic liver disease, which use the nodular lesions in the Proton density and T2–weighed magnetic resonance images on the gray level of the pixels in the region of interest.

Multimedia services based on broadband ISDN (B-ISDN) technology need a network architecture that satisfies the requirements of users, carriers, and vendors. This paper describes a new network architecture for B-ISDN service control and management based on INA. We list general requirements, and present implementation issues of INA. A network architecture and main components, which resolve implementation issues, is then proposed. We also describe a video-on-demand service based on our proposed architecture.

A method is presented for reconstructing the surface profile of a perfectly conducting rough surface boundary from the measurements of the scattered far-field. The proposed inversion algorithm is based on the use of the Kirchhoff approximation and in order to determine the surface profile, the Fletcher-Powell optimization procedure is applied. A number of numerical results illustrating the method are presented.

This paper describes a new feature extraction model (Active Model) which is extended from the active contour model (Snakes). Active Model can be applied to various energy minimizing models since it integrates most of the energy terms ever proposed into one model and also provides the new terms for multiple images such as motion and stereo images. The computational order of energy minimization process is estimated in comparison with a dynamic programming method and a greedy algorithm, and it is shown that the energy minimization process in Active Model is faster than the others. Some experimental results are also shown.

Based on a newly proposed notion of relational network, a novel learning mechanism for model acquisition is developed. This new mechanism explicitly deals with both qualitative and quantitative relations between parts of an object. Qualitative relations are mirrored in the topology of the network. Quantitative relations appear in the form of generalized predicates, that is, predicates that are graded in their validity over a certain range. Starting from a decomposition of binary objects into meaningful parts, first a description of the decomposition in terms of relational networks is obtained. Based on the description of two or more instances of the same concept, generalizations are obtained by first finding matchings between instances. Generalizing itself proceeds on two levels: the topological and the predicate level. Topological generalization is achieved by a simple rule-based graph generalizer. Generalization of the predicates uses some ideas from MYCIN. After successful generalization, the system attempts to derive a simple and coarse description of the achieved result in terms of near natural language. Several examples underline the validity of relational networks and illustrate the performance of the proposed system.

We introduce recurrent networks that are able to learn chaotic maps, and investigate whether the neural models also capture the dynamical invariants (Correlation Dimension, largest Lyapunov exponent) of chaotic time series. We show that the dynamical invariants can be learned already by feedforward neural networks, but that recurrent learning improves the dynamical modeling of the time series. We discover a novel type of overtraining which corresponds to the forgetting of the largest Lyapunov exponent during learning and call this phenomenon dynamical overtraining. Furthermore, we introduce a penalty term that involves a dynamical invariant of the network and avoids dynamical overtraining. As examples we use the Hnon map, the logistic map and a real world chaotic series that correspond to the concentration of one of the chemicals as a function of time in experiments on the Belousov–Zhabotinskii reaction in a well–stirred flow reactor.

We propose two methods to fuse auditory information and visual information for accurate sppech recognition. The first method fuses two kinds of information by using linear combination after calculating two kinds of probabilities by HMM for each word. The second method fuses two kinds of information by using the histogram which expresses the correlation of them. We have performed experiments comparing the proposed methods with the conventional method and confirmed the validity of the proposed methods.

This paper describes a segmentation method of liver structure from abdominal CT images using a three–layered neural network (NN). Before the NN segmentation, preprocessing is employed to locally enhance the contrast of the region of interest. Postprocessing is also automatically applied after the NN segmentation in order to remove the unwanted spots and smooth the detected boundary. To evaluate the performance of the proposed method, the NN–determined boundaries are compared with those traced by two highly trained surgeons. Our preliminary results show that the proposed method has potential utility in automatic segmentation of liver structure and other organs in the human body.

We report on an experimental hysteresis in the Hopfield networks and examine the effect of the hysteresis on some important characteristics of the Hopfield networks. The detail mathematic description of the hysteresis phenomenon in the Hopfield networks is given. It suggests that the hysteresis results from fully–connected interconnection of the Hopfield networks and the hysteresis tends to makes the Hopfield networks difficult to reach the global minimum. This paper presents a T–Model network approach to overcoming the hysteresis phenomenon by employing a half–connected interconnection. As a result, there is no hysteresis phenomenon found in the T–Model networks. Theoretical analysis of the T–Model networks is also given. The hysteresis phenomenon in the Hopfield and the T–Model networks is illustrated through experiments and simulations. The experiments agree with the theoretical analysis very well.

This paper presents a cost-effective network for very large ATM cross-connects. In order to develop it, we propose the delta network with expanded middle stages. This proposed network is the intermediate network between a nonblocking network and the delta network with respect to the cost of hardware and internal blocking probability. Using this network, we explore the tradeoff between the cost and internal blocking probability, and derive the optimum configuration under temporarily deviating traffic. Internal blocking occurs when input traffic temporarily deviates from its average value. However, we cannot evaluate the internal blocking probability by using conventional traffic models. In this paper, we adopt temporarily deviating traffic such that all traffic is described as the superposition of the paths which are defined by traffic parameters. As can easily be seen, the path corresponds to virtual path (VP) or virtual channel (VC). Therefore, we believe that our model describes actual traffic more exactly than conventional models do. We show that the optimum configuration is the proposed network whose expansion ratio γ=3 when the maximum number of paths that can be accommodated in one link is greater than 22. This network achieves the internal blocking probability of 10-10. As an example of this network, we show that the proposed network of size 7272 is constructed with only 40% of the hardware required by the nonblocking network.