The optical ATM transport networks are the key technology for B-ISDN which integrates wide variety of communication services. In a photonic ATM switch, electronically operated switching control drastically limits the total throughput. With this bottleneck of the control speed of the switch compared to the cell transmission speed, if the unit of switching operation can be made longer, the system throughput will be improved. This paper proposes the optical backbone network configuration to obtain traffic concentrating effect to construct a large switching unit called a multi-cell. In the backbone network applied the concept of virtual path set (VPS), a multi-cell is constructed by cells from all of ATM switches in a regional network connected with each own's cross connect. The multi-cell format in the case of two different network models is investigated in this paper. The average delay and average idle cells per multi-cell in the multi-cell MUX of an optical cross connect are evaluated by computer simulation as the performance of the multi-cell transfer scheme in the backbone network. Simulation results show that the multi-cell transfer scheme can be operated efficiently with traffic load of more than 0.5. This paper also proposes the configuration of multiplexing and demultiplexing module to assemble and disassemble the multi-cell.

This paper describes a model of kanji (Chinese characters) learning, called KASTAM (Knowledge Acquisition and STAbility Model), in order to develop a CAI system which supports a student to stabilize her/his kanji knowledge. KASTAM can handle kanji learning as complementary acquisition and stability processes of kanji knowledge.

Various approaches on optical network systems using wavelength-division multiplexing (WDM) technique have been proposed. It is difficult to make a scale of WDM network larger since a number of the optical wavelength which can be used is limited. In order to make easily larger scale of network, multi-hop WDM network have been proposed. We have studied 2-hop network: RookNet which has simple routing algorithm and high network throughput. Nodes in RookNet are divided into row groups and column groups and are placed in a mesh form. Packets are transferred between nodes over 1-hop or 2-hops. The 2-hop transfer means that a source node sends packets to a destination node via a relay node. When 2-hop traffic increases, relay processing time in a relay node is increasing. This is the reason that network throughput becomes low. To solve this problem is very important. In this paper, we show RookNet rearrangement algorithm which replaces location of node within group so as to decrease the 2-hop traffic and to maintain high network throughput. Proposed rearrangement algorithm can achieve improvement of 10 percent in terms of throughput. We also propose RookNet configuration which discriminates optical wavelength and subcarrier effectively in order to decrease the relay processing time.

Various authors have proposed probabilistic extensions of Valiant's PAC (Probably Approximately Correct) learning model in which the target to be learned is a (conditional) probability distribution. In this paper, we improve upon the best known upper bounds on the sample complexity of the parameter estimation part of the learning problem for distributions and stochastic rules over a finite domain with respect to the Kullback-Leibler divergence (KL-devergence). In particular, we improve the upper bound of order O(1/ε2) due to Abe, Takeuchi, and Warmuth to a bound of order O(1/ε). In obtaining our results, we made use of the properties of a specific estimator (slightly modified maximum likelihood estimator) with respect to the KL-divergence, while previously known upper bounds were obtained using the uniform convergence technique.

Most traditional directory systems use tree structure. This structure has revealed some problems, especially in a large system domain like distributed networking environment. These problems are severe constraints on object naming, no support to automatic sharing, and single name space. We proposed a new directory structure called MSD (Multi-Space Directory) to solve these problems. MSD exhibits several outstanding characteristics over tree directory, such as supporting both hierarchical and orthogonal classifications, supporting both vertical and horizontal automatic sharings, providing multiple working environment, and providing a better mechanism for protection and consistency. This paper first analyzed the problems of tree directory and new requirements for directory systems, then describes the structure of MSD in detail and shows its merits over tree directory.

We present a superconducting neural network which functions as an RS flip-flop. We employ a coupled-SQUID as a neuron, which is a combination of a single-junction SQUID and a double-junction SQUID. A resistor is used as a fixed synapse. The network consists of two neurons and two synapses. The operation of the network is simulated under the junction current density of 100 kA/cm2. The result shows that the network is operated as an RS flip-flop with clock speed capability up to 50 GHz.

This paper discusses some problems in Molecular Biology for which learning paradigms are strongly desired. We also present a framework of knowledge discovery by PAC-learning paradigm together with its theory and practice developed in our work for discovery from amino acid sequences.

A multiple instruction stream-multiple data stream (MIMD) computer is a parallel computer consisting of a large number of identical processing elements. The essential feature that distinguishes one MIMD computer family from another is the interconnection network. In this paper, we are concerned with a representative type of interconnection networks: the hypercube connected network. A family of regular graphs is presented as a possible candidate for the implementation of a distributed system and for fault-tolerant architectures. The symmetry of graphs makes it possible to determine message routing by using a simple distributed algorithm. A candidate having the same property is the hypercube connected network. Arbitrary data permutations are generally accomplished by sorting. For certain classes of permutations, however, this is, for many frequently used permutations in parallel processing such as bit reversal, bit shuffle, bit complement, matrix transpose, butterfly permutations used in FFT algorithms, and segment shuffles, there exist algorithms that are more efficient than the best sorting algorithm. One such class is the bit permute complement (BPC) class of permutations. In this paper, we, first, develop an algorithm to realize an arbitrary BPC permutation in hypercube connected networks. The developed algorithm in hypercube connected networks requires only 1 token memory register in each node. We next evaluate the ability to realize BPC permutations in these networks of an arbitrary size by estimating the number of required routing steps.

A self-induced-transparent (SIT) system that takes advantage of morphology dependent resonances (MDR's) in a Mie-sized microsphere doped with a resonant material is proposed. The present system is doubly resonant: one has microscopic origin (the two-level system), while the other has macroscopic origin (the MDR). In this geometry, owing to the feedback action of MDR's, the pulse area can be much expanded, and thus the electric-field amplitude of the incident pulse can be reduced substantially compared with the conventional one-way SIT propagation. Theoretical results that incorporate dephasing due to structural imperfections are shown.

The selection of an appropriate key search algorithm for a specific application field is an important issue in application systems development. This is because data retrieval is the most time-consuming part of many application programs. An automatic selection method for key search algorithms is presented in this paper. The methodology has been implemented in a system called KESE2 (KEy-SEarch ALgorithm SElection). Key search algorithms are selected according to the user's requirements through interaction with KESE2 which bases its inferences on an evaluation table. This evaluation table contains values rating the performance of each key search algorithm for the different searching properties, or characteristics. The selection algorithm presented is based on step by step reduction of unsuitable key search algorithms and searching properties. The paper also proposes assistance facilities that consist of both a support function and a program synthesis function. Experimental results show that the appropriate key search algorithms are effectively selected, and that the necessary number of questions asked, to select the appropriate algorithm, is reduced to less than half of the total number of possible questions. The support function is useful for the user during the selection process and the program synthesis function fully translates a selected key search algorithm into high level language in an average of less than 1 hour.

We are aming to develop a computer-based consultant system which helps novice computer users to achieve their task goals on computers through natural language dialogues. Our target is spoken Japanese. To develop effective methods for processing spoken Japanese, it is essential to analyze real dialogues and to find the characteristics of spoken Japanese. In this paper, we discuss the design problems associated with constructing a spoken dialogue database from the viewpoint of advisory dialogue collection, describe XMH (X-window-based electronic mail handling program) usage experiments made to collect advisory dialogues between novice XMH users and an expert consultant, and show the dialogue database we constructed from these dialogues. The main features of our database are as follows: (1) Our target dialogues were advisory ones. (2) The advisory dialogues were all related to the use of XMH that has a visual interface operated by a keyboard and a mouse. (3) The primary objective of the users was not to engage in dialogues but to achieve specific task goals using XMH. (4) Not only what the users said but also XMH operations performed by the users are included as dialogue elements. This kind of dialogue database is a very effective source for developing new methods for processing spoken language in multimodal consultant systems, and we have therefore made it available to the public. Based on our analysis of the database, we have already developed several effective methods such as a method for recognizing user's communicative intention from a transcript of spoken Japanese, and a method for controlling dialogues between a novice XMH user and the computer-based consultant system which we are developing. Also, we have proposed several response generation rules as the response strategy for the consultant system. We have developed an experimental consultant system by implementing the above methods and strategy.

A Hierarchical Cubic Network (HCN) is a hierarchical hypercube network proposed by Ghose. The HCN is topologically superior to many other similar networks, in particular, the hypercube. It has a considerably lower diameter than a comparable hypercube and is realized using almost half the number of links per node as a comparable hypercube. In this paper, we propose the shortest routing algorithm in HCN(n, n) and show that the diameter of HCN(n, n) with 22n nodes is n(n1)/31 which is about 2/3 of that of a comparable hypercube. We also propose the optimal routing algorithm in HCN(m, n) where mn and obtain that its diameter is n(m1)/31. Typical parallel algorithms run in HCN(m, n) with the same time complexity as a hypercube and the hypercube topology can be emulated with O(1) time complexity in it.

In this paper, we propose a method, called Kernel Hidden Unit Analysis, to reduce the network size. The kernel hidden unit analysis in composed of two principal components: T-component and S-component. The T-component transforms original networks into the networks which can easily be simplified. The S-component is used to select kernel units in the networks and construct kernel networks with kernel units. For the T-component, an entropy function is used, which is defined with respect to the state of the hidden units. In a process of entropy minimization, multiple strongly inhibitory connections are to be generated, which tend to turn off as many units as possible. Thus, some major hidden units can easily be extracted. Concerning the S-component, we use the relevance and the variance of input-hidden connections and detect the kernel hidden units for constructing the kernel network. Applying the kernel hidden unit analysis to the symmetry problem and autoencoders, we perfectly succeeded in obtaining kernel networks with small entropy, that is, small number of hidden units.

This paper discusses two approaches to statistical multiplexing: rate envelope multiplexing, allowing resource sharing with small delays for low peak rate connections, and rate sharing, based on the use of large multiplexer buffers to ensure high link utilization for high speed data traffic. We argue that the weighted fair queueing scheduling algorithm provides an efficient means for combining both kinds of multiplexing in the B-ISDN. A feasible implementation known as Virtual Spacing is outlined. We illustrate the flexibility of the proposed scheme by showing how different service categories could be provided.

Calls using different media which require different transfer quality will arrive at ATM networks. Therefore it is important to develop a method for allocating network resources efficiently to individual calls by judging admission of calls. Various call admission control schemes have been already proposed, and these schemes assume that users specify values of traffic descriptors when they originate calls. However, it is sometimes difficult for users to specify these values accurately. This paper proposes a new ATM call admission control scheme based on cell transfer state monitoring which does not require that users specify values of traffic descriptors in detail when they originate calls. In this proposed scheme, the acceptance or rejection of calls is judged by comparing the monitored cell transfer state value with a threshold prepared in advance. This threshold must be adjusted according to changes in the characteristics of traffic applied to ATM networks. This is one of the most serious problems in the control scheme based on the monitoring of cell transfer state. Herein, this paper proposes neural network application to the control scheme in order to resolve this problem and improve performance. In principle, the threshold can be adjusted automatically by the self-learning function of the neural network, and the control can be maintained appropriately even if the characteristics of traffic applied to ATM networks change drastically. In this paper, the effectiveness of the application of a neural network is clarified by showing the configuration of this proposed control scheme with the neural network, a method for deciding various parameter values needed to implement this control scheme, and finally the results of a performance evaluation of the control scheme. Inputs required by the neural network are also discussed.

The Distributed Queue Dual Bus protocol (DQDB) has been adopted as the metropolitan area network (MAN) standard by IEEE802.6 committee. Recently, the unfairness problem in the DQDB protocol, by which head stations benefit, has been pointed out. Although a fair bandwidth distribution among the stations is obtained by adding the so-called bandwidth balancing mechanism into the DQDB protocol (DQDB/BB), the DQDB/BB protocol leaves a portion of the available bandwidth unused, and it takes a considerable amount of time to converge to fair channel assignment. In this study, to overcome the drawbacks in DQDB and DQDB/BB, we introduce a new media access control protocol which is based on assigning each station a level according to some traffic information such as the queueing length, delay time etc. Only the station with the highest level is allowed to transmit. Through the operation of level assignment or the choice of level function, the transmission can be easily controlled in a distributed manner. This protocol is simple compared with DQDB/BB and can be implemented in the DQDB architecture. The simulation results show that the new protocol obtains not only fair throughput regardless of the distance between the stations, but also fair delay performance. In addition, the new protocol can easily provide preempty priority service through level assignment. The new protocol converges to fair distribution of the channel in the time required for only one or two round-trips. This is very fast compared with the DQDB/BB protocol.

This paper describes reverse modulation carrier recovery with a tank-limiter for Offset QPSK (OQPSK) burst signals. Acquisition performance is discussed taking into account hardware implementation errors in the carrier recovery circuit. The results indicate hardware implementation errors cause a significant recovered carrier phase error during BTR (Bit Timing Recovery) of OQPSK burst signals. A phase error reduction technique by modifying the BTR code for OQPSK burst signals is proposed to improve the acquisition performance. Computer simulation and hardware experiments confirmed its improvement. The performance of a prototype OQPSK burst demodulator using the proposed carrier recovery scheme is also presented.

A graph G is said to be universal for a family F of graphs if G contains every graph in F as a subgraph. A minimum universal graph for F is a universal graph for F with the minimum number of edges. This paper considers a minimum universal graph for the family Fkn of graphs on n vertices with path-width at most k. We first show that the number of edges in a universal graph Fkn is at least Ω(kn log(n/k)). Next, we construct a universal graph for Fkn with O(kn log(n/k)) edges, and show that the number of edges in a minimum universal graph for Fkn is Θ(kn log(n/k)) .

Switched-capacitor chaotic neurons fabricated in a full-custom integrated circuit are used to investigate the behavior of 2- and 3-neuron chaotic neural networks. Various sets of parameters are used to visualize the dynamical responses of the networks. Hysteresis of the network is also demonstrated. Lyapunov exponents are approximated from the measured data to characterize the state of each neuron. The effect of the finite length of data and the rounding effect of data acquisition system to the computation of Lyapunov exponents are briefly discussed.

We study the correspondence between problem descriptions and requirements specification documents derived from them. Based on the results of this investigation, a model that integrates the problem space and the requirements specification space is developed. This integration is based on a semantic network representation. We also propose a model of the requirements elicitation process that is consistent with our empirical studies of traceability in requirements documents. In this process, analysts derived requirements specifications from incomplete and ambiguous problem descriptions given by customers, identify missing information, completed it, and then decide the system boundaries that define which part of the problem descriptions to implement as the target system. The model can be used to complete problem descriptions given by customers and determine the system boundaries.