Hisashi Shimosaka

Distributed probabilistic model-building genetic algorithm

In this paper, a new model of Probabilistic Model-Building Genetic Algorithms (PMBGAs), Distributed PMBGA (DPMBGA), is proposed. In the DPMBGA, the correlation among the design variables is considered by Principal Component Analysis (PCA) when the offsprings are generated. The island model is also applied in the DPMBGA for maintaining the population diversity. Through the standard test functions, some models of DPMBGA are examined. The DPMBGA where PCA is executed in the half of the islands can find the good solutions in the problems whether or not the problems have the correlation among the design variables. At the same time, the search capability and some characteristics of the DPMBGA are also discussed.

Distributed Workflow Management System based on Publish-Subscribe Notification for Web Services

In recent years, Grid technologies have been standardized based on Web service specifications. Of these specifications, the WS-Resources Framework and WS-Notification have attracted a great deal of attention. This paper focuses on scientific applications integration on the wide area network. We propose and implement a new distributed workflow management system called the “Application Igniting System.” This system is based on the publish-subscribe notification defined by the WS-Notification specification and realizes a flexible and loosely coupled workflow control by introducing some Web services, which handle message exchange. By applying to a typical bioinformatics workflow, we concluded that the overhead time related to message exchange is very short.

Structural optimization by real-coded probabilistic model-building GA

In this paper, a probabilistic model-building genetic algorithm (PMBGA) is applied to structural optimization problems. PMBGA has high searching ability but it sometimes converges to the local minimum. To avoid this problem, the concept of distributed GA is applied to PMBGA. To deal with constraints, the penalty function and pulling back methods are also applied to PMBGA. Using the proposed methods, a truss structure is designed to minimize its volume as a numerical example. Through the numerical example, the comparison between PMBGA and conventional DGA shows the effectiveness of PMBGA. The penalty function and pulling back methods are also effective in the example.

Optimization Problem Solving System using GridRPC

The invention provides HSV antigens that are useful for the prevention and treatment of HSV infection. Disclosed herein are epitopes confirmed to be recognized by T-cells derived from herpetic lesions. T-cells having specificity for antigens of the invention have demonstrated cytotoxic activity against cells loaded with virally-encoded peptide epitopes, and in many cases, against cells infected with HSV. The identification of immunogenic antigens responsible for T-cell specificity provides improved anti-viral therapeutic and prophylactic strategies. Compositions containing antigens or polynucleotides encoding antigens of the invention provide effectively targeted vaccines for prevention and treatment of HSV infection.

Offspring generation method using delaunay triangulation for real-coded genetic algorithms

To design crossover operators with high search ability in real-coded Genetic Algorithms, it will be efficient to utilize both information regarding the parent distribution and the landscape of the objective function. Here, we propose a new offspring generation method using Delaunay triangulation. The proposed method can concentrate offspring in regions with a satisfactory evaluation value, inheriting the parent distribution. Through numerical examples, the proposed method was shown to be capable of deriving the optimum with a smaller population size and lower number of evaluations than Simplex Crossover, which uses only information of the parent distribution.

Satisfactory design of cogeneration system using genetic algorithm

This paper introduces the optimum design of co-generation system (CGS) using the genetic algorithm (GA). CGS is the energy reusing system which generates more than two energies from one energy source. To design CGS, the types of machines and load scheduling should be determined. However, the optimum design of CGS is too complicated even for the expert. One of the solutions for this problem is using GA. GA is the optimization model imitating evolution of life. If the coding of the problems is proper, GA can be applicable to many problems. However, proper coding for the problems is difficult, especially for CGS, because it has three different design variables which consist of integer values and real values. To discuss the effective coding, this paper considers four models. First is simplest coding model. Second is two-step optimization model with integer coding. Third is two-step optimization model with the integer coding and the penalty method. Last is three-step optimization model with the integer. As a result of the experiments, three-step optimization model could achieve the higher energy efficiency design of CGS than the expert

Comparison of pulling back and penalty methods for constraints in DPMBGA

To solve real-world problems by genetic algorithms (GAs), GAs that have a strong searching capability are needed. In this paper, distributed probabilistic model building genetic algorithm (DPMBGA) is applied to solve the problems. The DPMBGA is an extended algorithm of probabilistic model building GA (PMBGA) and it also has a strong searching capability. In real world problems, constraints often exist. As such, mechanisms that can treat the constraints should be added to the GAs. Two mechanisms for treating constraints are the penalty method and pulling back method. The DPMBGA with penalty method and pulling back method is applied to truss structural optimization problems. Through a simulation, the searching capability and efficiency of the pulling back method and penalty method are discussed. From the discussion, it is concluded that the pulling back method can derive the solutions even if the problem is difficult. Compared to the penalty method, the number of individuals that violate the constraints is smaller in the pulling back method.

Voronoi model-building genetic algorithm

This paper proposes the Voronoi model-building genetic algorithm (VMBGA), which is one of real-coded GAs. In the VMBGA, a Voronoi model is constructed using with Voronoi diagrams. Because of this mechanism, the distribution of offspring can adapt to the landscape of the objective function by changing the Voronoi model. Through the some standard test functions, the effectiveness of the VMBGA is examined. It is clarified that the VMBGA has higher searching ability than the UNDX-m, which is one of the typical real-coded GAs. Additionally, the distribution of the offspring is also discussed.