Maximum adjacency (MA) ordering has effectively been applied to graph connectivity problems by Nagamochi and Ibaraki. We show an application of MA ordering to the maximum flow problem to get a new polynomial-time algorithm and propose its scaling versions that run in O (mnlogt7) time, where TO is the number of arcs, n the number of vertices, and U the maximum capacity. We give computational results, comparing our algorithms with those of Goldberg-Tarjan and Dinitz, to show behaviors of our proposed algorithms.

View full abstract

This paper deals with a traffic grooming and light-path routing problem in WDM-ring net-works, where the number of light-paths to be used for each traffic demand is restricted. With a network augmentation by adding a dummy node and a set of dummy arcs, we formulate the comprehensive problem as a mixed Integer Programming problem. Owing to the computational complexity, it is hard to find an optimal solution for a large-scale network within a reasonable computation time. Hence, we develop a heuristic algorithm to solve it with easy. Our heuristic generates good feasible solutions within a few second in computation time even for large-scale networks. The computational experiments show that the performance of the proposed heuristic is satisfactory in both the speed and the quality of the solutions generated.

View full abstract

Data envelopment analysis (DEA) is an important analytical tool for estimating the technical efficiency of decision-making units However, input allocative efficiency is less frequently investigated in the DEA framework because the measure of input allocative efficiency-derived as the ratio of cost efficiency to Farrell input technical efficiency-may be biased because of slack in the constraints that define the technology or because of differences in input prices paid by firms. We extend the DEA methodology by proposing a new measure of input allocative efficiency based on the direct and indirect output quasi-distance functions An empirical illustration of the new and existing methods for measuring input allocative efficiency is offered for a sample of U.S. banks operating in 1994-95

View full abstract

This paper deals with an improvement method for the Workflow Management Systems (WMSs) based on the reallocation of flows using the Genetic Programming (GP) and its applications Current WMSs are usually used in a static manner and WMSs are designed based on conventional experiences It is also necessary to modify the system by observing the failed tasks We propose a method for the improvement and reconfiguration of WMSs based on the GP It is expected that the scheme of approximation of chaotic dynamics by using the GP is extended to the WMS design In our method, each activity on WMSs is assumed to be a function having several inputs, and the combination of these functions and sequences of the functions are improved by using the GP. Links combining nodes are assumed to be flows in networks The syntactical validity of the individuals are checked by using a kind of counter for the parse tree obtained in the prefix representation If the validity is not satisfied, the individual is removed from the pool of individuals In the improvement, the evaluation function is composed of the the completion of underlying objectives as well as conventional measures such as costs and waiting time We apply the crossover operations to the individuals possessing relatively higher fitness in the reproduction phase The mutation operation is utilized especially to determine the configuration of split operations by reintroducing some diversity By using the result of simulation, sequences of database access obtained by the flows in WMSs are examined whether they included at least the control sequence imposed on the underlying WMS realization The method is applied to the improved design of WMSs and the extension to the hierarchical system design

View full abstract

This paper deals with a datum search game, where a target reveals his position (datum point) at a certain time (datum time) and a pursuer begins the search for the target by distributing his searching effort some time later The target might move in the diffusive fashion from the datum point to evade his pursuer His motion is restricted by its continuity in a two-dimensional space and constraints on its energy and maximum speed The pursuer distributes searching effort to detect the evader under constraints on the amount of effort. A payoff is assumed to be the summation of searching effort weighted by the probability distribution of the target In the previous paper, we formulated the problem as a single-stage two-person zero-sum game on continuous space and continuous time and proposed an upper bound and a lower bound for the value of the game. This paper extends the result and proposes an approximation for the value, noting that a constant-speed motion is crucial for the target

View full abstract

This paper considers a FIFO single-server queue with service interruptions and multiple batch Markovian arrival streams. The server state (on and off), the type of arriving customers and their batch size are assumed to be governed by a continuous-time Markov chain with finite states. To put it more concretely, the marginal process of the server state is a phase-type alternating Markov renewal process, the marginal arrival process is a batch marked Markovian arrival process, and they may be correlated Further, service times of arriving customers are allowed to depend on both their arrival stream and the server state on arrival. For such a queue, we derive the vector joint generating function of the numbers of customers from respective arrival streams. Further assuming discrete phase-type batch size distributions, we establish a numerical algorithm to compute the joint queue length distribution at a random point in time. Finally, we show some numerical examples and examine the impact of system parameters on the queue length distribution

View full abstract

We consider an optimal stopping problem with a discrete time stochastic process where a criterion is a threshold probability. We first obtain the fundamental characterization of an optimal value and an optimal stopping time as the result of the classical optimal stopping problem, but the optimal value and the optimal stopping time depend upon a threshold value We also give the properties of the optimal value with respect to threshold value These are applied to a secretary problem, a parking problem and job search problems and we explicitly find an optimal value and an optimal stopping time for each problem

View full abstract

A baseball game between teams consisting of non-identical players is modeled using a Markov chain, taking into account the number of runs by which the home team leads. Using the Markov model the probability of winning in any state in the course of a game is calculated directly by solving a set of over one million simultaneous equations. This approach makes it possible to obtain the optimal pinch hitting strategy under the 'Designated Hitter' rule by applying dynamic programming to this model. We demonstrate this method using a match based on the line-ups of the Anaheim Angels and the Oakland Athletics in the American League of Major League Baseball. We show how this approach may help to determine when to use a pinch hitter and how much the probability of winning increases if the optimal strategy is followed.

View full abstract

[in Japanese]

View full abstract