Eitaro Aiyoshi

Keio University, Edo, Tokyo, Japan

Are you Eitaro Aiyoshi?

Claim your profile

Publications (91)29.11 Total impact

  • Minoru Kanemasa · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: Modern heuristic optimization algorithms developed in '90s have been a particular focus of attention because of their simplicity, easy software implementation, and moreover, the interesting phenomena that their performance emerged from the interactions among the particles. In this paper, we see that we can get emergent performance as an optimization algorithm by increasing the number of particles on Evolution Strategy. Considering that, we try to increase the interactions among the particles in order to get better performance. We define parameter tuning rule designing as an optimization problem, and use Genetic Programming to find those for Evolution Strategy. In addition, we evaluate the generated tuning rules using statistical tests and several benchmarks to verify that the proposed methods and the generated rules are effective ones.
    No preview · Article · Mar 2015 · IEEJ Transactions on Electronics Information and Systems
  • Yuji Koguma · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: In Recent years, a paradigm of optimization algorithms referred to as “meta-heuristics” have been gaining attention as a means of obtaining approximate solutions to optimization problems quickly without any special prior knowledge of the problems. Meta-heuristics are characterized by flexibility in implementation. In practical applications, we can make use of not only existing algorithms but also revised algorithms that reflect the prior knowledge of the problems. Most meta-heuristic algorithms lack mathematical grounds, however, and therefore generally require a process of trial and error for the algorithm design and its parameter adjustment. For one of the resolution of the problem, we propose an approach to design algorithms with mathematical grounds. The approach consists of first constructing a “framework” of which dynamic characteristics can be derived theoretically and then designing concrete algorithms within the framework. In this paper, we propose such a framework that employs two following basic strategies commonly used in existing meta-heuristic algorithms, namely, (1) multipoint searching, and (2) stochastic searching with pseudo-random numbers. In the framework, the update-formula of search point positions is given by a linear combination of normally distributed random numbers and a fixed input term. We also present a stability theory of the search point distribution for the proposed framework, using the variance of the search point positions as the index of stability. This theory can be applied to any algorithm that is designed within the proposed framework, and the results can be used to obtain a control rule for the search point distribution of each algorithm. We also verify the stability theory and the optimization capability of an algorithm based on the proposed framework by numerical simulation.
    No preview · Article · Feb 2015 · IEICE Transactions on Fundamentals of Electronics Communications and Computer Sciences
  • Minoru KANEMASA · Eitaro AIYOSHI
    [Show abstract] [Hide abstract]
    ABSTRACT: Crossover is one of the most known nature inspired operations in heuristic optimization algorithms. It was traditionally inspired by the evolution of species, and it is well known for the capability of solving discrete optimization problems, which use integer valued vector. However, in recent years, algorithms such as Differential Evolution use it to solve continuous optimization problem, regardless of the fact that this usage is apart from the way species evolve using DNA sequence. This kind of crossover operation in continuous space creates new points in axis-wise directions, thus it is known that the performance of those algorithms using continuous crossover have different performance when we rotate the coordinate of an optimization problem. This is because uniform crossover is not a rotation invariant operation. In this paper, we consider of using rotationally invariant crossover called hypersphere crossover. However, since this crossover may not adopt to ill-conditioned problem with fixed radius, we propose scaling parameter and its tuning rule to change the radius of the hypersphere to compensate for the problem. We compare our proposal with traditional uniform crossover, and other rotation invariant crossover operations using many benchmarks. We use pair-wises ranked t-test to statistically verify the advantage of our proposal.
    No preview · Article · Jan 2015
  • Minoru Kanemasa · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: In recent years, improvements in processing power have allowed the application of optimization methods to complicated large optimization problems. Among these methods, heuristic optimization techniques such as particle swarm optimization (PSO) have been a particular focus of attention because of their simplicity, performance, and easy software implementation. However, there is no solid theoretical foundation for analyzing the convergence of these algorithms, and in practice, their rate of convergence is often determined by the choice of parameters. For this reason, the algorithm's parameters must be tuned appropriately for each new optimization problem we want to solve, and in some cases the parameters must be varied as the algorithm is updated. In this paper, we combine a feedback element as an algorithm tuner with an original algorithm; the resulting algorithm is applied to the optimization problem in question, and we use genetic programming (GP) to generate tuning rules to automatically tune the algorithm, according to its current state, as the algorithm is updated. More specifically, we adopt PSO as a heuristic optimization method, and we augment PSO by using GP as a meta-algorithm to solve the learning problem of automatically generating tuning rules for the parameters in the PSO algorithm. This leads to the proposed method for generating parameter tuning rules to solve optimization problems more efficiently. © 2014 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.
    No preview · Article · Jul 2014 · IEEJ Transactions on Electrical and Electronic Engineering
  • Takashi Okamoto · Yuya Hanaoka · Eitaro Aiyoshi · Yoko Kobayashi
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we consider the use of a multi‐objective optimization method in order to obtain a preferred solution for the buffer material optimal design problem in the high‐level geological disposal of radioactive waste. The buffer material optimal design problem is formulated as a constrained multi‐objective optimization problem. Its Pareto optimal solutions are distributed evenly over the entire feasible region. Hence, we develop a search method to find a preferred solution easily for a decision maker from the Pareto optimal solutions, which are distributed evenly and broadly. In the preferred solution search method, a technique for visualization of a Pareto optimal solution set using a self‐organizing map is introduced into the satisficing trade‐off method, which is an interactive method for obtaining a Pareto optimal solution that satisfies a decision maker. We confirm the effectiveness of the preferred solution search method in the buffer material optimal design problem. © 2014 Wiley Periodicals, Inc. Electr Eng Jpn, 187(2): 17–32, 2014; Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/eej.22634
    No preview · Article · Apr 2014 · Electrical Engineering in Japan
  • Yuji Koguma · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: Recently, a paradigm of optimization method called meta-heuristic (MH) has attracted interest for its applicability. However, the most part of MHs have insufficient mathematical background in their dynamics. Some of MHs adopt pseudo-random numbers generated by a computer in their algorithm so that they are called “stochastic optimization methods”. As for stochastic optimization methods, the positions of search points are able to be regarded as stochastic variables which are distributed according to a certain probability distribution. In this paper, we perform an analysis of search points distribution for DE dynamics based on maximum entropy method.
    No preview · Article · Jan 2014 · IEEJ Transactions on Electronics Information and Systems
  • Takashi Okamoto · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: In this study, we transform an optimization problem into a game problem by multiple decision makers. This transformation is called gamification of the optimization problem. Then, we discuss correspondence of the dynamics to search the rational solution of the game problem with the multi-point type optimization method to search the optimal solution of the original problem. Specifically, we propose a condition to compose objective functions of multiple decision makers who interfere with each other from the objective function of the original problem so that equivalence of the Nash equilibrium solution of the game problem to the optimal solution of the original problem is guaranteed. Then, correspondence of the gradient dynamics to solve the game problem with the multi-point type optimization method is discussed. We also show some numerical examples in order to explain the gamification of the optimization problem.
    No preview · Conference Paper · Oct 2013
  • Y. Koguma · E. Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: Particle swarm optimization (PSO), a meta-heuristic global optimization method, has attracted special interest for its simple algorithm and high searching ability. The updating formula of PSO involves coefficients with random numbers as parameters to enhance diversification ability in searching for the global optimum. However, the randomness makes stability of the searching points difficult to analyze mathematically, and the users need to adjust the parameter values by trial and error. In this paper, stability of the stochastic dynamics of PSO is analyzed mathematically and an exact stability condition taking the randomness into consideration is presented with an index called the "statistical eigenvalue," which is a new concept for evaluating the degree of stability of PSO dynamics. The accuracy and effectiveness of the proposed stability discrimination using the presented index are certified in numerical simulation for simple examples.
    No preview · Article · Jan 2012
  • Takashi Okamoto · Yuya Hanaoka · Eitaro Aiyoshi · Yoko Kobayashi
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we consider a multi-objective optimization method in order to obtain a preferred solution for the buffer material optimal design problem in the high-level radioactive wastes geological disposal. The buffer material optimal design problem is formulated as a constrained multi-objective optimization problem. Its Pareto optimal solutions are distributed evenly on whole bounds of the feasible region. Hence, we develop a search method to find a preferred solution easily for a decision maker from the Pareto optimal solutions which are distributed evenly and vastly. In the preferred solution search method, the visualization technique of a Pareto optimal solution set using the self-organizing map is introduced into the satisficing trade-off method which is the interactive method to obtain a Pareto optimal solution that satisfies a decision maker. We confirm the effectiveness of the preferred solution search method in the buffer material optimal design problem.
    No preview · Article · Jan 2012 · IEEJ Transactions on Electronics Information and Systems
  • Kenichi Muranaka · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we present a new type of hybrid methods for global optimization with Particle Swarm Optimization (PSO) and Differential Evolution (DE), which have attracted interests as heuristic and global optimization methods recently. Concretely, “p-best solutions” as the targets of PSO's particles are actuated by DE's evolutional mechanism in order to promote PSO's global searching ability. The presented hybrid method works effectively because PSO acts as a local optimizer and DE plays a role as a global optimizer. To evaluate performance of the hybridization, our method is applied to some benchmarks and is compared with the separated PSO and DE. Through computer simulations, it is certified that the proposed hybrid method performs fairy better than their separated algorithm.
    No preview · Article · Jan 2012 · IEEJ Transactions on Electronics Information and Systems
  • Source
    Eitaro Aiyoshi · Atsushi Maki · Takashi Okamoto
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a Nash equilibrium model that applies continuous time replicator dynamics to the analysis of oligopoly markets. The robustness of the proposed simple Nash equilibrium model under the simultaneous constraints of allocation of product and market share using a simulation method to derive an optimal solution for production decisions by rival firms in oligopoly markets is tested by changing profit and cost function parameters, as well as the initial production values and market shares of the firms examined in this study. The effects of differences in conjectural variation and initial allocation of market share on the convergent values are considered, particularly in the case of corner solutions. This approach facilitates the understanding of the robustness of attaining equilibrium in an oligopoly market.
    Full-text · Article · Mar 2011 · Mathematics and Computers in Simulation
  • Takashi Okamoto · Eitaro Aiyoshi · Kenichiro Hamada
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we propose a new global optimization model which is described as discrete time-varying inertial gradient dynamics. The proposed model has a autonomous damping term adjustment structure which takes the search history into consideration. In the proposed model, its dynamics destabilizes autonomously when its search point approaches to the best point in the search history. Thus, stagnation of search around the neighborhood of the best point is restrained, in consequence, global search is continued. Furthermore, in this paper, we analytically explain a characteristic of the search trajectory generated by the inertial gradient system. The characteristic is the implementation of intensive and diverse search in the attracting region for the inertial gradient system. Consequently, we propose a multi-point type search model in which search points are attracted to a promising region, where objective function value is small, by a coupling structure in order to make the attracting region the promising region. We confirm effectiveness of the proposed model through numerical experiments using several benchmark problems.
    No preview · Article · Jan 2011
  • Yoshinao Ishii · Takashi Okamoto · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: Particle Swarm Optimization (PSO), which has attracted attention as a global optimization method in recent years, has a drawback in that sustainable search cannot be performed until the end of computation due to its strong convergence trend. In this paper, in order to realize a sustainable search in PSO, the improved PSO using concepts of particle ages and digenesis is proposed. In the new PSO, parameters in the update formula are degenerated and a stagnant particle is erased if it loses activity, and then a new search point in which large parameter values are assigned. In addition, information regarding the elite point of all searching points until the current time is reflected to new points in next generation. The effectiveness of the improved method is confirmed through applications to benchmark problems.
    No preview · Article · Jan 2011 · IEEJ Transactions on Electronics Information and Systems
  • Yuji Koguma · Atsuro Furusawa · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: It is known that the evolutional computing methods hybridized with local search methods, which are called Memetic Algorithm, are efficient as global optimization methods. The memetic algorithms are divided into two classes of hybridization strategies, which are called Baldwinian and Lamarchian types. This paper is concerned with computational considerations for these two types of hybridization strategies in a case when Particle Swarm Optimization is used as evolutional algorithm from the standpoint of global optimization methods. Especially, applications to optimization problems with constraints described by disconnected plural sets, which are difficult to solve by usual methods, are considered.
    No preview · Article · Jan 2011
  • Kazuaki Masuda · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: We propose a method for solving optimal price decision problems for simultaneous multi-article auctions. An auction problem, originally formulated as a combinatorial problem, determines both every seller's whether or not to sell his/her article and every buyer's which article(s) to buy, so that the total utility of buyers and sellers will be maximized. Due to the duality theory, we transform it equivalently into a dual problem in which Lagrange multipliers are interpreted as articles' transaction price. As the dual problem is a continuous optimization problem with respect to the multipliers (i.e., the transaction prices), we propose a numerical method to solve it by applying heuristic global search methods. In this paper, Particle Swarm Optimization (PSO) is used to solve the dual problem, and experimental results are presented to show the validity of the proposed method.
    No preview · Article · Jan 2011 · IEEJ Transactions on Electronics Information and Systems
  • Kazuaki Masuda · Kenzo Kurihara · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, a method for solving min-max problems, especially for finding a solution which satisfies “min-max = max-min” condition, by using a modified particle swarm optimization (PSO) algorithm, is proposed. According to recent development in computer science, multi-point global search methods, most of which are classified into evolutionary computation and/or meta-heuristic methods, have been proposed and applied to various types of optimization problems. However, applications of them to min-max problems have been scarce despite their theoretical and practical importance. Since direct application of evolutionary computation methods to min-max problems wouldn't work effectively, a modified PSO algorithm for solving them is proposed. The proposed method is designed: (1) to approximate the minimized and maximized functions of min-max problems by using a finite number of search points; and, (2) to obtain one of “min-max = max-min” solutions by finding the minimum of the maximized function and the maximum of the minimized function. Numerical examples demonstrate the usefulness of the proposed method.
    No preview · Conference Paper · Jan 2011
  • Kazuaki Masuda · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: An autoassociative memory which is modeled as the standard recurrent neural network (N.N.) is capable of storing multiple patterns and subsequently recalling one of them in response to an input signal. However, we found in our recent trials that it can't always recall correct patterns accurately. In this paper, we demonstrate such phenomena by numerical examples and identify the cause of memorization errors. We also propose an immediate solution to memorize correct patterns without fail by storing extra patterns at the same time.
    No preview · Conference Paper · Jan 2011
  • Sodo Hiraoka · Takashi Okamoto · Eitaro Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: Particle Swarm Optimization (PSO), which has attracted a great deal of attention as a global optimization method in recent years, has a drawback in that its continuous search based on its excellent dynamic characteristics can not be executed stably until the end of computation due to its much strong convergence trend. In this paper, we propose “Repetitive Search Guideline” which differs from a common guideline in the improved methods which have ever been proposed and by which the continuous search of PSO is achieved without lack of PSO's excellent dynamic characteristics due to the repetitive search in a promise area where objective function value is expected to be small. We consider four improved methods based on the proposed guideline, and then, their effectiveness are confirmed through applications to 100 variables multi-peaked benchmark problems.
    No preview · Article · Nov 2010 · IEEJ Transactions on Electronics Information and Systems
  • K. Masuda · K. Kurihara · E. Aiyoshi
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper proposes a penalty method for solving nonlinear optimization problems with inequalities by the particle swarm optimization (PSO) algorithm. The proposed method is not only very simple but also useful. One should only search for the global solution of a series of unconstrained minimization problems simply by a standard PSO algorithm. It does not require to check the feasibility of search points during the search. Moreover, it is shown that the global best solution gets feasible as the penalty parameter is increased to a sufficiently but finitely large value. The proposed method is verified by numerical experiments to famous benchmark problems.
    No preview · Conference Paper · Nov 2010
  • Eitaro Aiyoshi · Kazuaki Masuda
    [Show abstract] [Hide abstract]
    ABSTRACT: On the basis of market fundamentalism, new types of social systems with the market mechanism such as electricity trading markets and carbon dioxide (CO2) emission trading markets have been developed. However, there are few textbooks in science and technology which present the explanation that Lagrange multipliers can be interpreted as market prices. This tutorial paper explains that (1) the steepest descent method for dual problems in optimization, and (2) Gauss-Seidel method for solving the stationary conditions of Lagrange problems with market principles, can formulate the mechanism of market pricing, which works even in the information-oriented modern society. The authors expect readers to acquire basic knowledge on optimization theory and algorithms related to economics and to utilize them for designing the mechanism of more complicated markets.
    No preview · Article · Jan 2010 · IEEJ Transactions on Electronics Information and Systems