著者

後藤 裕介 森田 裕之 白井 康之 市川 尚 濱田 直希 原田 智広

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.13, no.1, pp.23-39, 2022 (Released:2022-09-09)

参考文献数

27

---

In recent years, evidence-based policy-making (EBPM) has been called for to accommodate diverse stakeholders when local governments formulate new policies. Social simulation allows virtual observation of changes in social conditions resulting from various alternatives in policy-making. However, there has not been a generic social simulation for designing subsidy payment policies that can be used in various situations. The Evolutionary Computation Competition 2021 (EC Comp 2021), an optimization competition that has been held since 2017 and intends to promote interaction between industry and academia, asked participants to design subsidy payment policies with social simulation. EC Comp 2021 newly formulates a generic social simulation framework for designing subsidy payment policies. This social simulation estimates the effects of subsidy payment policies in response to changes in household economic conditions based on economic shock scenarios using statistically valid data on the residents in a city. This paper gives a detailed explanation of the subsidy payment design problem with the social simulation in EC Comp2021. This paper explains the participants’ optimization methods and their results, accompanied by a brief analysis of their results, and discusses the characteristics of the optimization problem.

著者

濱田 直希 於保 俊 谷垣 勇輝 原田 智広 能島 裕介

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.12, no.3, pp.112-124, 2021 (Released:2022-01-18)

参考文献数

23

---

The Evolutionary Computation Competition (EC-Comp) is an optimization competition launched in 2017 to promote real-world applications of evolutionary computation and interaction between industry and academia. For 2017—2019, the competition has focused on continuous optimization problems in the manufacturing and aerospace industries. With the aim of exploring new areas of applications, EC-Comp2020 focused on "Designing Random Numbers to Entertain Game Players" in the game industry. Random numbers used in video games are usually generated by general-purpose pseudo random number generators, such as Mersenne Twister and Xorshift. However, these mathematically unbiased random numbers often make game players feel biased (sometimes even deliberately chosen), causing strong frustration. It is known that humans have various biases toward probabilistic events, and unbiased random numbers seem rather biased to game players. This competition asked to design a random number sequence that makes game players feel unbiased (but actually biased). This paper describes the definition of the random number design problem for entertaining game players in EC-Comp2020. This paper also explains the participants' optimization methods, accompanied with brief analysis on their results.

著者

能島 裕介 高木 英行 棟朝 雅晴 濱田 直希 西原 慧 高玉 圭樹 佐藤 寛之 桐淵 大貴 宮川 みなみ

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.13, no.1, pp.1-9, 2022 (Released:2022-07-13)

参考文献数

4

---

This paper is a report on Open Space Discussion (OSD) held in Evolutionary Computation Symposium 2021. The purpose of OSD is to share and discuss problems at hand and future research targets related to evolutionary computation. Discussion topics are voluntarily proposed by some of the participants, and other participants freely choose one to join in the discussion. Through free discussions based on the open space technology framework, it is expected that participants will have new research ideas and start some collaborations. This paper gives the concept of OSD and introduces six topics discussed this year. This paper also shows the responses to the questionnaire on OSD for future discussions, collaborations, and related events.

著者

益富 和之 永田 裕一 小野 功

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.6, no.1, pp.1-12, 2015 (Released:2015-04-28)

参考文献数

28

---

This paper proposes a novel evolution strategy for noisy function optimization. We consider minimization of the expectation of a continuous domain function with stochastic parameters. The proposed method is an extended variant of distance-weighted exponential evolution strategy (DX-NES), which is a state-of-the-art algorithm for deterministic function optimization. We name it DX-NES for uncertain environments (DX-NES-UE). DX-NES-UE estimates the objective function by a quadratic surrogate function. In order to make a balance between speed and accuracy, DX-NES-UE uses surrogate function values when the noise is strong; otherwise it uses observed objective function values. We conduct numerical experiments on 20-dimensional benchmark problems to compare the performance of DX-NES-UE and that of uncertainty handling covariance matrix adaptation evolution strategy (UH-CMA-ES). UH-CMA-ES is one of the most promising methods for noisy function optimization. Benchmark problems include a multimodal function, ill-scaled functions and a non-C2 function with additive noise and decision variable perturbation (sometime called actuator noise). The experiments show that DX-NES-UE requires about 1/100 times as many observations as UH-CMA-ES does on well-scaled functions. The performance difference is greater on ill-scaled functions.

著者

原田 圭 廣安 知之 日和 悟

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.9, no.2, pp.75-85, 2018 (Released:2018-10-10)

参考文献数

26

---

MOEA/D decomposes a multiobjective optimization problem into a set of single objective subproblems. When there are a few differences in difficulty of each objective function, it can obtain widely-spread and uniformly-distributed solutions. However, in real-world problems, the complexities of the objective functions are often heterogeneous. In this case, each subproblem of the MOEA/D has different difficulty so that the spread and uniformity of the population is deteriorated because the search direction in the objective space tends to be biased into the feasible region which is easily explored. To overcome this issue, an adaptive weight assignment strategy for MOEA/D is proposed in this paper. In the proposed method, the subproblems are divided into some groups and the convergence speed is estimated for each group and utilized as the metric of the difficulty of the subproblems. Moreover, the weight vectors of easy subproblem groups are modified to bias their search into the subproblem group with higher difficulty. Our proposed method is validated on the region-of-interests determination problem in brain network analysis whose objective functions have heterogeneous difficulties. The experimental results showed that our method worked better than the conventional weight assignment strategy in MOEA/D.

著者

金崎 雅博 千葉 一永 北川 幸樹 嶋田 徹

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.6, no.3, pp.137-145, 2015 (Released:2015-12-18)

参考文献数

19

---

With the multi-combustion technology, the combustion in a hybrid rocket engine (HRE) can be temporarily stopped via oxidizer throttling. In this paper, two types of HREs, one with multi-combustion technology and the other without, are compared to investigate the effects of multi-combustion on the flight performance of launch vehicles (LVs). Non-dominated Sorting Genetic Algorithm-II (NSGA-II) which was a multi-objective evolutionary algorithm (MOEA) was applied to solve the design problems using real-number coding and the Pareto ranking method. To investigate the effects of the multi-combustion on flight performance of LV with HRE, three design problems were considered. The first case was the maximization of the flight altitude and the minimization of the gross weight. The second case was the minimization of the maximum acceleration and the minimization of the gross weight. The final case was the maximization of the flight downrange and the minimization of the gross weight. Many non-dominated solutions were obtained by NSGA-II, and a trade-off was observed between the two objective functions in each case. MOEA results were visualized using a parallel coordinate plot. According to the exploration result, it was found that the multi-combustion of HRE was effective to reduce the maximum acceleration. Such ability could be expected to reduce the shock load to payloads carried by the LV with HRE.

著者

千葉 一永 金崎 雅博

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.4, no.1, pp.28-37, 2013 (Released:2013-05-17)

参考文献数

22

---

Design informatics, which is the efficient design methodology, has three points of view. The first is the efficient exploration in design space using evolutionary-based optimization methods. The second is the structurization and visualization of design space using data mining techniques. The third is the application to practical problems. In the present study, the influence of the difference among the seven pure and hybrid optimization methods for design information has been investigated in order to explain the selection manner of optimization methods for data mining. The practical problem of a single-stage hybrid rocket is picked up as the present design object. A functional analysis of variance and a self-organizing map are employed as data mining techniques in order to acquire the global design information in dasign space. As a result, mining result depends on not the number of generation (i.e. convergence) but the optimization methods (i.e. exploration space). Consequently, the optimization method with diversity performance is the beneficial selection in order to obtain the global design information in design space.

著者

中山 惠太 白川 真一 矢田 紀子 長尾 智晴

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.3, no.2, pp.12-21, 2012 (Released:2012-06-11)

参考文献数

12

---

Non-photorealistic rendering (NPR), a research about non-photorealistic images is a major field of research in image processing. Painterly rendering is a method that creates artistic images based on photo images and important in NPR. Recently, painterly rendering methods using evolutionary algorithm are studied. Those studies have intended to optimize the process of creating artistic images by using evolutionary algorithm. Most of those studies have focused on generating and placing strokes as a painting operation. On the other hand, some researchers proposed painterly rendering methods using existing art images. They have created painting images which have unique colors and textures of existing art images, called “painting style”. We propose a new method to create artistic images based on photo images and existing art images by using Genetic Algorithm (GA). Our method operates putting the “patches” on a canvas image repeatedly as a painting operation. We generate the “patches” by copying a part of the existing art images and put them on the canvas image in mutation of GA. We exchange pixels in the same region of two canvas images in crossover of GA. In the process of optimization, our method brings the canvas image close to the photo images. Our method evolutionarily creates the painting images which have the painting styles of existing art images.

著者

渡邉 真也 奥寺 将至

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.5, no.3, pp.32-44, 2014 (Released:2014-11-26)

参考文献数

13

被引用文献数

1

---

Nurse scheduling problem (NSP) is one of the most popular constraint satisfaction problems and its importance is very high to maintain high quality medical services and avoid staff's work overload in real world. Even though there are some commercial scheduling software for calculating optimal assignment of shifts and holidays to nurses, these haven't met user's demand yet in the points of computational time and diversity of candidate schedulings. Since it is widely known that the constraints of NSP can be categorized into two main types; nursing quality and staff's quality of life, NSP can be treated as two-objective optimization problem. In this paper, a new approach for NSP is proposed. The proposed approach is based on evolutionary multi-criterion optimization (EMO) and its main features are high search performance and derivation of plural different candidate solutions. This research is collaboration with System Bank Co.,Ltd. and its mission is to improve an existing optimization engine.To investigate the characteristics and effectiveness of the proposed approach, the proposed is applied to three different benchmark problems which have characters and difficulty levels. The results of numerical examples provided that the performance of the proposed approach is overwhelming in comparison with the existing engine in every problems. Also, each mechanism's works of the proposed approach can be apparent through numerical examples.

著者

岸 祐希 アリヤリ アタフォン 金崎 雅博

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.12, no.3, pp.137-147, 2021 (Released:2022-01-18)

参考文献数

32

---

In this study, a multi-fidelity approach was developed based on the efficient global optimization (EGO) and integrated with multi-additional sampling. The developed approach was more efficient than the conventional multi-fidelity approach when applied to design problems. The effectiveness of the proposed approach was demonstrated by solving two test problems (a test problem in Van Valedhuizen’s test suite and a test problem with a convex Pareto front) before applying the approach to real-world problems. As a demonstration of solving real-world problem, we solved two objective airfoil design problems for a small unmanned airplane. The objective functions were the drag coefficient (for flight efficiency) and the thickness at the 75% chord position (for structural strength and manufacturability). The results of the test problems revealed that the proposed approach obtained more non-dominant solutions near the theoretical Pareto front than those obtained by the Original optimization approach at the same iteration number of EGO loop; this is because the proposed approach obtained more additional samples than the Original optimization approach (multi-objective multi-fidelity EGO without multi-additional sampling) per additional sampling loop. A comparison of the accuracies of surrogate models based on the proposed approach and the Original optimization approach using leave-one-out cross validation suggested that, depending on the optimization problem, one of the two approaches can yield greater accuracy. The airfoil design results, as well as the test problems, revealed that the proposed approach can obtain several better solutions than those obtained by the Original optimization approach when the number of iterations of additional sampling was the same between both approaches. The hypervolume in the proposed approach also increases more rapidly than that in the Original optimization approach.

著者

鳥山 直樹 小野 景子

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.9, no.2, pp.32-40, 2018 (Released:2018-06-01)

参考文献数

18

---

In this paper, we present an efficient sampling method for a multimodal and high-dimensional distribution. For sampling from a high-dimensional distribution, DE-MC, which is based on the Markov chain Monte Carlo(MCMC) methods, has been proposed. It showed good performance in sampling from any probability distribution based on constructing a Markov chain that has the desired distribution. However, DE-MC has inherent difficulties in sampling from a multimodal distribution. To overcome this problem, we incorporate a replica exchange method into DE-MC and propose a replica exchange resampling DE-MC method (reRDE-MC) based on sampling importance resampling to improve its performance. The proposed method is evaluated by using three types of distributions with multimodal and high dimensions as artificial data. We verified that the proposed method can sample from a multimodal and highdimensional distribution more effectively than by a conventional method. We then evaluated the proposed method by using financial data as actual data, and confirmed that the proposed method can capture the behavior of financial data.

著者

太田 恵大 佐藤 寛之

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.10, no.2, pp.22-32, 2019 (Released:2020-02-13)

参考文献数

28

---

For air-conditioning systems in office buildings, it is crucial to reduce power consumption while maintaining office workers' thermal comfort. This paper proposes a simulation-based evolutionary multi-objective air-conditioning schedule optimization system for office buildings. In the proposed system, a target office building is modeled and simulated by EnergyPlus building simulator which is one of the practical simulators widely used in the building construction field. To obtain the temperature schedules which dynamically change the temperature setting over time, we use an improved multi-objective particle swarm optimization algorithm, OMOPSO, to simultaneously optimize the thermal comfort of office workers in the building and the power consumption of the air-conditioning system. Experimental results show that the proposed system can obtain temperature schedules better than the conventional schedule with constant temperature settings from viewpoints of both the thermal comfort and the power consumption. Also, we show experimental results that the multi-objective search in the proposed system acquires better temperature schedules than single objective particle swarm optimization and differential evolution algorithms using ε-constraint method as one option of single objective optimization approaches. Furthermore, we show that OMOPSO obtains temperature schedules widely approximating the optimal tradeoff between the thermal comfort and the power consumption compared with other evolutionary multi-objective optimizers, NSGA-II, NSGA-III, MOEA/D-DE.

著者

石渕 久生 半田 久志

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.1, no.1, pp.15-22, 2011-09-30 (Released:2011-09-30)

参考文献数

94

被引用文献数

1

---

Evolutionary Computation (EC) has been receiving growing attention in the last two decades. EC has been utilized practically not only in computer science but also in other areas. This paper introduces recent advances in EC by referring to the impact factors of EC journals, the acceptance rates of three major EC conferences, and the most cited Japanese papers at CiNii. In addition, active research areas in EC such as evolutionary multi-objective optimization, swarm intelligence and real-world applications are briefly described.

著者

小平 剛央 釼持 寛正 大山 聖 立川 智章

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.8, no.1, pp.11-21, 2017 (Released:2017-08-04)

参考文献数

27

被引用文献数

1

---

We propose two optimization benchmark problems with actual engineering design features of car-body structural development using response surface method. The first is a single-objective optimization problem of weight minimization. The second is the multi-objective optimization problem of weight minimization and number of common thickness parts maximization. These benchmark problems have key feature of real-world design problem, i.e. many variables and many constraint conditions. Furthermore, it is a discrete variable optimization problem. We present optimization results using multi-island genetic algorithm (single-objective optimization) and NSGA-II (multiobjective optimization). We also present optimization result of the benchmark problem where the discrete variables are treated as real variables. The benchmark problem is available on http://ladse.eng.isas.jaxa.jp/benchmark.

著者

大山 聖

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.9, no.2, pp.86-92, 2018 (Released:2018-10-10)

参考文献数

24

---

Evolutionary computation competition 2017 was held in December 9, 2017 in conjunction with evolutionary computation symposium 2017. It was confirmed that evolutionary algorithms can discover good designs of the design optimization problem of vehicle structures provided by Mazda motor company. Nine teams participated in the single-objective optimization division and eleven teams in the multiobjective optimization division. Prof. Shinya Watanabe's team from Muroran Institute of Technology won in the single-objective optimization division, Prof. Isao Ono's team from Tokyo Institute of Technology won in the multi-objective optimization division. The industrial use special prize was awarded to Dr. Tomohiro Harada's team from Ritsumeikan University. In the single-objective design optimization division, the groups using evolution strategies found good Pareto-optimal solutions. In the multiobjective optimization division, the groups who found good Pareto-optimal designs studied characteristics of the benchmark problem very much and implemented the most suitable optimization algorithm. Mazda benchmark problem has many severe constraints and thus feasible design space is strictly limited. Some teams used special techniques such as ε constraint method. Current result indicated that balance between search in feasible region and infeasible region may be important for constrained design optimization problems.

著者

西田 昂平 秋本 洋平

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.8, no.2, pp.61-74, 2017 (Released:2017-12-01)

参考文献数

17

---

The population size, i.e., the number of candidate solutions per iteration, is the only parameter for the covariance matrix adaptation evolution strategy (CMA-ES) that needs to be tuned depending on the ruggedness and the uncertainty of the objective function. The population size has a great impact on the performance of the CMA-ES, however, it is prohibitively expensive in black-box scenario to tune the population size in advance. Moreover, a reasonable population size is not constant during the optimization. In this paper, we propose a novel strategy to adapt the population size. We introduce the evolution path in the parameter space of the Gaussian distribution, which accumulates successive parameter updates. Based on the length of the evolution path with respect to the Fisher metric, we quantify the accuracy of the parameter update. The population size is then updated so that the quantified accuracy is kept in the constant range during search. The proposed strategy is evaluated on test functions including rugged functions and noisy functions where a larger population size is known to help to find a better solution. The experimental results show that the population size is kept as small as the default population size on unimodal functions, and it is increased at the early stage of the optimization of multimodal functions and decreased after the sampling distribution is concentrated in a single valley of a local optimum. On noisy test functions, the proposed strategy start increasing the population size when the noise-to-signal ratio becomes relatively high. The proposed strategy is compared with the CMA-ES and the state-of-the-art uncertainty handling in the CMA-ES, namely UH-CMA-ES, with a hand-tuned population sizes.

著者

上村 健人 木下 峻一 永田 裕一 小林 重信 小野 功

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.4, no.1, pp.1-12, 2013 (Released:2013-03-02)

参考文献数

18

被引用文献数

1

---

This paper proposes a new framework of real-coded genetic algorithms (RCGAs) for the multi-funnel function optimization. The RCGA is one of the most powerful function optimization methods. Most conventional RCGAs work effectively on the single-funnel function that consists of a single big-valley. However, it is reported that they show poor performance or, sometimes, fail to find the optimum on the multi-funnel function that consists of multiple big-valleys. In order to remedy this deterioration, Innately Split Model (ISM) has been proposed as a framework of RCGAs. ISM initializes an RCGA in a small region and repeats a search with the RCGA as changing the position of the region randomly. ISM outperforms conventional RCGAs on the multi-funnel functions. However, ISM has two problems in terms of the search efficiency and the difficulty of setting parameters. Our proposed method, Big-valley Explorer (BE), is a framework of RCGAs like ISM and it has two novel mechanisms to overcome these problems, the big-valley estimation mechanism and the adaptive initialization mechanism. Once the RCGA finishes a search, the big-valley estimation mechanism estimates a big-valley that the RCGA already explored and removes the region from the search space to prevent the RCGA from searching the same big-valley many times. After that, the adaptive initialization mechanism initializes the RCGA in a wide unexplored region adaptively to find unexplored big-valleys. We evaluate BE through some numerical experiments with both single-funnel and multi-funnel benchmark functions.

著者

伊庭 斉志

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.1, no.1, pp.23-31, 2011-09-30 (Released:2011-09-30)

参考文献数

30

---

Genetic Programming (GP) has a relatively short but exciting history. This interesting filed has been steadily growing. However, there had been a conflict between GPers and GA people at the earlier stage of research emergence. It dramatically seems to have demonstrated the systematic abuse of the peer review process and of unethical behavior, intellectual dishonesty, and scientific misconduct. In this paper we show such historical anecdotes, e.g., the massacre of GP papers in ML95, proposed ICGA95 arrangements, and FOIA requests etc. We give some lessons from the above historical events, hoping that history never repeats itself in our community.

著者

福島 信純 永田 裕一 小林 重信 小野 功

出版者

進化計算学会

雑誌

進化計算学会論文誌 (ISSN:21857385)

巻号頁・発行日

vol.4, no.2, pp.57-73, 2013 (Released:2013-09-11)

参考文献数

19

---

The natural evolution strategies (NESs) is a family of iterative methods for black-box function optimization. Instead of directly minimizing an objective function, NESs minimizes the expectation of the objective function value over an arbitrary parametric probability distribution. In each iteration, NESs updates parameters of the distribution by using an estimated natural gradient of the expectation of the objective function value. Exponential NES (xNES) is an effective method of NESs that uses the multivariate normal distribution as the probability distribution. Since the shape of a normal distribution can take the form of a rotated ellipse in the solution space, xNES shows relatively good performance for ill-conditioned and non-separable objective functions. However, we believe that xNES has two problems that cause performance degradation. The first problem is that the spread of normal distribution tends to shrink excessively even if the distribution does not cover a (local) optimal point. This will cause premature convergence. The second problem is that the learning rates for the parameters of distribution are not appropriate. The learning rates depend only on the dimension of objective function although they should be designed depending on all the factors that influence the precision of natural gradient estimation. Moreover, they are set to small values for preventing the premature convergence and these results in too slow convergence speed even if the distribution covers the optimal point. In order to remedy the problems of xNES, we propose a new method of NESs named the distance-weighted exponential natural evolution strategy (DX-NES). On several benchmark functions, we confirmed that DX-NES outperforms xNES and that DX-NES shows better performance than CMA-ES on the almost all functions.