Differential Evolution (DE) is one of the effective metaheuristics for solving unconstrained optimization problems. Constraint Handling Technique (CHT) is needed to extend DE to constrained optimization. Feasibility Rule (FR) is one of the typical CHT. FR addresses constraints by using a simple rule that considers the objective function and constraint violation when comparing search individuals. However, since the solutions with small constraint violation, i.e., feasible solutions, are preferentially selected, the set of search individuals may be biased toward feasible regions and the improvement of the objective function value may stagnate. This paper overcomes this challenge by proposing a DE that uses a superior infeasible solution in an external archive. The external archive in the proposed DE stores search individuals that are superior in both objective function value and constraint violation and utilize them to generate mutant individuals. Finally, we verify the effectiveness of the proposed method using a benchmark problem where the feasible region is a convex set.

This paper proposes an adaptive PBI for massively parallel MOEA/D in a distributed memory environment. Massively parallelization in a distributed memory environment effectively speeds up evolutionary multi-objective optimization algorithms for practical application problems. On the other hand, when MOEA/D is divided for parallelization by focusing on the reference vector in the objective function space, the T-neighbor is divided and the problem that the solution distribution becomes sparse near the boundary of the divided region arises. Here, we propose a method to improve the problem that the T-neighbor is divided and the solution distribution becomes sparse by adaptively controlling the penalty value in the PBI function according to the distance from the reference vector using a distribution function such as Laplace distribution. The effectiveness of the proposed method is shown by comparison with execution using a single CPU.

While there has been many new developments in multiobjective evolutionary algorithms, they have not been applied or investigated in clustering problems. In this paper, ten different unsupervised clustering techniques applying different MOEA (SPEA2, IBEA, MOEA/D and MOEA/GLU), PSO (QPSO and BBPSO) and Fuzzy approaches are experimented on ten public datasets. The rationale to apply MOEA is to increase the exploitation capabilities of clustering techniques to further refine the cluster solutions found by fuzzy or PSO clustering. The aim is to investigate in the performance of different types of MOEA applications in clustering, determining whether MOEA Fuzzy clustering outperform MOEA PSO variants. Overall, MOEA/D BBPSO was found to produced the best results. It outperformed MOEA Fuzzy techniques, having tested on datasets with high number of classes, that are imbalanced and/or overlapping classes. IBEA Fuzzy clustering was found to produce the worst results. MOEA/D clustering was found to perform better than other MOEA techniques. In this work, we showed that MOEA/D BBPSO clustering produced the best results on challenging datasets. It was able to use MOEA/D to deepen its exploitation capability while benefiting from the exploratory ability of BBPSO when clustering challenging datasets.

This paper proposes an update method to determine the reference point and nadir point when speeding up multi-objective evolutionary optimization algorithm based on decomposition (MOEA/D) with parallel and distributed processing in a many-core environment. We have already proposed a migration method that uses a virtual overlapping zone between partitions. Compared with simply partitioning the objective space or applying the standard island migration model, this method is shown to be effective in improving the hypervolume (HV) value, increasing diversity, and reducing regions where non-dominated solutions do not exist between partitions (sparse areas). However, this proposed method could not completely deal with the sparse areas' problem in the early stage of solution searching. To resolve this problem, in addition to the virtual overlap method, we propose here a method that switches the update interval for determining the reference point and nadir point by taking into account the progress of solution searching. To evaluate our newly proposed method, we compare both convergence and diversity on single-core MOEA/D and parallel MOEA/D using a two-objective constrained knapsack problem, and show that our method is effective not only in reducing sparse areas in the early stage but also in improving diversity and increasing the HV value.

Radiometric calibration with the Moon (called the lunar calibration) is a promising method for evaluating the performance of instruments onboard satellites. In particular, the lunar calibration can provide inflight radiometric calibration opportunities for nano/microsatellites after the launch without any special equipment. In this study, we have conducted more than one year of Moon observations with a microsatellite named RISESAT. The performance of an onboard multispectral sensor OOC is investigated by using the ROLO and SELENE/SP Moon models. Our lunar calibration revealed the dependence of the sensor sensitivity on the instrument temperature, and no sensitivity degradation (< 1 %) can be confirmed for all four channels after the correction of the temperature dependence. Lunar calibration also revealed the bluing trend in the OOC’ s inter-band ratio, which indicates a 15% alternation in the sensitivity at maximum. This bluing trend was validated by the Railroad Valley Playa observations as comparing the Landsat-8/OLI data.

Symbolic execution is a powerful technique for automating software testing to detect many types of errors such as memory errors and assertion violations. However, it encounters the problem of path explosion, and by using only assertions, it still lacks the capability of going deep into checking the functional correctness of a path based on corresponding formal specifications. To address these problems, we propose a specification-based incremental testing method with symbolic execution, called SIT-SE, providing a much more rigorous way to automatically check the functional correctness of all the discovered program paths, by introducing theorems (instead of assertions) for path correctness and branch sequence coverage algorithm for guiding a moderate path exploration. Compared with Hoare logic for proving the correctness of an entire program, a theorem in the SIT-SE is made for verifying the correctness of a program path. The proposed method carefully treats the relationship between a path condition and the specification in a theorem to restrict the monotonous path exploration, whereas traditional concolic testing methods roughly use one test data to determine the path correctness by assertions during long path searching. We use a classic case to demonstrate how the method works and conduct an experiment to evaluate the performance of both the proposed method and the commonly used well-known concolic testing tool KLEE. The experimental results show that our method SIT-SE is effective and outperforms KLEE in detecting faulty paths based on specifications.

This paper proposes a method for many-core-based large-scale parallel and distributed computation of MOEA/D, a decomposition-based evolutionary multi-objective optimization algorithm. Standard parallel MOEA/D on many-core environments provides fast execution time, but uniformity and diversity of the Pareto front may be lost. To avoid this problem, we propose a method of defining a virtual overlapping zone between partitions and selecting individuals for mating and migration by evaluating individual populations in this area using weight vectors of adjacent partitions. Using a two-objective constrained knapsack problem for evaluation, we compare the proposed method with standard single-core execution, no-migration parallel MOEA/D, and parallel MOEA/D with standard migration, and show that the proposed method is effective in improving diversity in solution searching while shortening execution time and increasing the accuracy of solution searching.

This paper proposes a novel voice adaptation method that we applied to interactive activities such as games where source and target data are unaligned. Conventional methods have seen the use of probabilistic models or more recently, Deep Neural Networks. Common for most methods is that they require multiple subjects to train in conjunction, thus voice adaptation is not practical to be used in commercial applications. We propose a method which convert audible frequencies to light spectrum simple RGB color format, and not comparing sound signal similarities, but rather likeness in color. The comparison is done using multi-objective optimization which considers raw and normalized frame colors as two separate objectives to be evaluated, respectively audible and spectral structure. The distance for the objectives is used to select an ideal output frame. Finally, prosodic information such as speech intensity is translated from measured input values onto the designated output frame. The method is evaluated using MOS, ABX, performance benchmark and lastly implemented into the Unity3D game engine as a proof of concept. Results show good sound quality and high performance with little output fragmentation.

The rapid development of safety-critical systems eagerly demands for the formal verification or enormous testing to ensure the correctness of programs. Testing-based formal verification with symbolic execution (TBFV-SE), a technique that integrates both formal verification and testing, taking the advantages of both methods without so much cost to meet such need. In order to further pinpoint the problematic positions in the code when TBFV-SE gives the incorrect program paths, we describe a fault localization algorithm to do this work through making use of the information provided by TBFV-SE. A case study demonstrates that our algorithm can more effectively and efficiently shrink the scale of suspicious lines of the code.

A recent trend of multi-objective evolutionary algorithms is to approximate the Pareto optimal front with high accuracy by increasing the size of the population. On the other hand, the cconventional method NSGA-II algorithm widely used in multi-objective optimization ranks solutions using non-dominant sorting that increase in computational complexity proportional to the population size squared. In engineering applications, the increased execution time can be a problem. Also, even if standard techniques for parallel processing of genetic manipulation are applied to multi-objective evolutionary computation, it is difficult to speed up while maintaining the accuracy of the solution search. This paper proposes the distributed NSGA-II which performs a hierarchical non-dominated sorting in a many-core environment, and migration that shares the extreme solutions of the latest Pareto optimal solutions among all cores. In the evaluation, using the test functions provided by the NSGA-II source code and the constrained knapsack problem with two-objective and three- objective, we describe that our proposed method contributes to the improvement of solution diversity search and for enhancing solution accuracy and speeding up.

As the current specification-based testing (SBT) face some challenges in test case generation for regression testing, we propose a new method for efficient test case generation that combines formal specifications with genetic algorithm (GA). This method mainly reforms formal specifications though GA to generate inputs that can kill as many as possible mutants of the target program under test. For case study, two classic examples are presented to demonstrate the efficiency of this method. The result shows that this method is able to help efficiently generate useful test cases to uncover all the program mutants, which contributes to further maintenance of software.

To reduce the computational cost of particle swarm optimization (PSO) methods, research has begun on the use of Graphics Processing Units (GPUs) to achieve faster processing speeds. However, since PSO methods search based on a global best value, they are hampered by the frequent need for communication with global memory. Even using a standard PSO that uses a local best value does not solve this problem. In this paper, we propose a virtual global best method that speeds up computations by defining a time-delayed global best as a virtual global best in order to reduce the frequency of communication with low-speed global memory. We also propose a method that combines decomposition-based multi-objective PSO (MOPSO/D) with a virtual global best method to speed up multi-objective particle swarm optimization by running it in parallel while maintaining search accuracy, and we demonstrate the effectiveness of this approach by using a number of unimodal/multimodal single objective benchmark test functions and three classical benchmark test functions with two objectives.

In this paper, we compared three different types of avatar design (“Body”, “Hand + Arm”, and “Hand only”) for the augmented reality remote instruction system in terms of usability. The result showed that the usability of the remote instruction system with full body avatar has a higher usability. In addition, participants felt more easily to track the full body avatar than the avatar with hand only. However, concerning the understandability of the instruction, there was no difference between three designs.

The bare bones particle swarm optimization algorithm is a useful method for the optimization problems. Each individual particle has been given memories to recorded its personal best position. The best of all personal best positions is recorded as the global best position by the particle swarm. A Gaussian distribution is used to control the behavior of the particles according to the personal and the global best position. However, this iterative pattern weak at multimodal problems. Particles are easy to be trapped in the local minimums. To cross this shortcoming, the dynamic reconstruction bare bones particle swarm optimization algorithm (DRBBPSO) is proposed in this work. The dynamic reconstruction strategy is used to enhance the global search ability of the particle swarm. Numbers of elite particles are selected to reconstruct the particle swarm. To verify the performance of the proposed algorithm, a set of comprehensive benchmark functions are used in the experiments. Also, several swarm-based algorithms including the standard bare bone particle swarm optimization algorithm are used in the control group. The experimental results confirmed the searching ability of the DRBBPSO.

This work proposes a decomposition-based algorithm, CMOEA/D-DMA, for constrained many-objective optimization. For constrained multi-objective optimization, the TNSDM algorithm using the directed mating selecting infeasible solutions having better objective values than feasible ones as parents was proposed and verified its effectiveness on several test problems. However, since TNSDM uses the non-dominated sorting, its search performance deteriorates when the number of objectives is increased. For many-objective optimization, the decomposing objective space is a promising approach, and MOEA/D is known as its representative algorithm. However, since the conventional MOEA/D maintains only one solution for each weight vector, feasible solutions are preferred rather than infeasible ones. Infeasible solutions are just discarded even if they have better objective values than maintained feasible solutions and can provide clues for the search. For solving constrained many-objective optimization problems, in this work, we propose CMOEA/D-DMA combining MOEA/D with the Directed Mating and Archives of infeasible solutions. The directed mating in CMOEA/D-DMA selects useful infeasible solutions having better scalarizing function values than feasible ones as parents and maintains them in archives. The experimental results using continuous mCDTLZ and discrete knapsack problems with many-objectives show that the proposed CMOEA/D-DMA achieves higher search performance than the conventional TNSDM, MOEA/D, and NSGA-III.

This paper proposes a nonparallel voice adaptation, a form of speech processing, which utilizes a voice profile with a mean dataset, paired with auxiliary sets generated from a stylized quantization from a single recording where only the target voice subject is required to train beforehand aimed towards real-time applications. Conventional methods suffer from complexified training where two or more speakers must be trained in parallel, thus impractical for actual real-time uses. Our quantized data is used to produce sound-indexes, a signature for each sound-frame for comparing speakers. Prosody is sustained by translating the input's power in the current frame into the corresponding output frame real-time. The design was tested with an English voice profile and the rate of matching output and frames were measured to determine a factor of fragmentation, which is the degree of unsuccessful segments. Results showed diverse fragmentation, complementing the input length. Basic input ranged in the very good 5-6% fragmentation range, while convoluted input up to 26%. The average matching time was 176 milliseconds. Future work focuses on improving performance and decreasing fragmentation.

Bare bones particle swarm optimization (BBPSO) algorithm, a swarm intelligence algorithm, is famous for its easy applying and parameter-free. That is why its principles and applications have been studied by a lot of scholars in recent years. However, quickly losing the diversity of the swarm still causes the premature convergence in the iteration process. Hence, a pair-wise bare bones particle swarm optimization (PBBPSO) algorithm is proposed in this paper to balance the exploration and exploitation. Moreover, a separate iteration strategy is used in pair-wise operator to enhance the diversity of the swarm. Also, to verify the performance of the proposed algorithm, a set of well-known nonlinear benchmark functions are used in the experiment. Furthermore, severe variants of BBPSO and some other evolutionarily algorithms are also evaluated on the same functions as the control group. Finally, the experiment result and statistical analysis confirm the performance of PBBPSO with nonlinear functions.

The bare bones particle swarm optimization (BBPSO) is a population-based algorithm. The BBPSO is famous for easy coding and fast applying. A Gaussian distribution is used to control the behavior of the particles. However, every particle learning from a same particle may cause the premature convergence. To solve this problem, a new hierarchical bare bones particle swarm optimization algorithm is proposed in this work. Three random particles are placed in one group and exchanging information during the iteration process. And a hierarchical method is used in every group. Therefore, the swarm gains an increasing of diversity and more chances to escape from the local optimum. Moreover, a mutated structure for the local group is presented in this paper. To verify the ability of the proposed algorithms, a set of well-known benchmark functions are used in the experiment. Also, to make the experiment more persuasive, several evolutionary computation algorithms are applied to the same functions as the control group. The experimental results show that the proposed algorithms perform well in the test functions.

This paper proposes a new modified Bare Bones Particle Swarms algorithm employed Differential Evolution strategy (BBPSO-DE) for large scale optimization. To the best of the authors' knowledge, a number of modified Particle Swarm Optimization algorithms have been suggested to solve large scale problems. However, the performance of these algorithms are influenced by their control parameters. The Bare Bones Particle Swarms Optimizer (BBPSO) is a simple and parameter-free algorithm where the positions of particles are sampled from Gaussian distribution. Although BBPSO is extensively used in literature, it also suffers premature convergence problem. In this paper, several strategies are integrated into BBPSO to alleviate premature convergence. These manipulations are mainly including using ring neighbourhood topology, introducing particle's local best position and centroid position of all particles to scale the standard deviation. If a particle's local best position is equal to personal best position, incorporated mutation and crossover operators of DE strategy is adopted in modified BBPSO, namely BBPSO-DE. Experimental results on test functions up to 1000 dimensions show that BBPSO-DE is a very competitive algorithm for solving both unimodal and multimodal problems. Furthermore, one key advantage of proposed BBPSO-DE is that there is no need to specify any control parameters that are often required in traditional stochastic algorithms.

This paper proposes a two-step based real-time voice adaptation system in the field of speech processing. Step one combines recording and pre-processing to construct a voice profile. Secondly is the real-time raw input of the voice's adaptation to a target voice. The fact that individual voices' structure are habitually varying, we suggest a method for converting into a comparable format. The new method is called abstract normalization which cuts the voice data into smaller sounds and generate an abstracted, simplified version of the data using a level of abstraction along with parameter fitting. The normalized data is used to generate a sound-index which is a sequence hash that represents the data in a simpler fashion. The indices are used to compare different sounds/voices for adaptation. This effectively transforms the speech-related challenges into a search problem rather than a biometric one. To assess the approach, voice profile data are compared against each other as a method to verify the sound-index. Ultimately, a real-time voice input using alternating levels of abstraction is run against a Norwegian voice profile. The degree of adaptation success is measured in percentage, and experimental results show that while accuracy is not yet excellent, the concept was validated.

To avoid premature convergence by maintaining diversity in the population, a new algorithm, Fitness Predator Optimizer (FPO), is proposed based on the concepts of predators. However, the slow convergence speed in the early iteration becomes the bottleneck to restrict the improvement of the FPO. In this paper, a modified dynamic virtual team for Fitness Predator Optimizer is proposed with the aim of accelerating the early convergence rate and improving the global searching capability for multimodal problems. The virtual dynamic team is a self-organization team that according to the nearest neighbour clustering. Each individual automatically is denoted as the team leader. The main function of the team leader is to obtain the neighbour team's best experience. It is worth noting that one of a team members may belong to another group concurrently. The expansion of individual best experience and the team best experience is carried out by this kind of multiple memberships. Consequently a paralleled exchanging information system is built on the modified dynamic virtual team to strengthen on the global searching capability during the early convergence process. Then an improved Dynamic Fitness Predator Optimizer (DFPO) with the modified virtual team is proposed in this paper. We extensively evaluated the DFPO and compared it with FPO, Dolphin Partner Optimization (DPO) and Standard Particle Swarm Optimization (SPSO), with respect to both running time and global solution capability. Experimental results show that the proposed DFPO could demonstrate the desirable performance for multimodal problems.

Credit erasure processing refers to the process of canceling corresponding items from a list of accounts receivable by checking against a detailed statement when receiving accounts receivable and other payments. In credit erasure processing, it is sometimes necessary to perform the laborious task of searching for credited items based solely on billing data and payments received. In this paper, we define credit erasure processing as a large-scale subset sum problem, and propose a solution based on a genetic algorithm (GA). In particular, we propose improving the search precision by incorporating a local search method that takes Hamming distances into consideration. To this end, we compare the standard GA with the result of adding a simple local search to the standard GA. This is done using a set of data where the numbers of digits in the billed quantities are roughly the same, another set of data where the numbers of digits are more varied, and a set of real data. As a result, we show that the proposed method works effectively when the numbers of digits in the billed quantities are roughly the same or when the fitness score is has a somewhat higher value.

A major problem with most of swarm intelligent algorithms in multimodal optimization is premature convergence (PC), which results in great performance loss and sub-optimal solutions. To avoid premature convergence by maintaining diversity in the population, many kinds of optimization algorithms are proposed. However, to the best of our knowledge, few of the swarm intelligent techniques focus on the individual competition. The development of individual competition plays an important role of the diversity conservation in the population because it could increase individual independent consciousness and reduce the rapid social collaboration process. In this paper, a new algorithm, Fitness Predator Optimizer (FPO), is proposed based on the conceptions of predators. In an FPO system, all of the individuals are seen as predators. Each of the individuals is depicted only by its position. Then the individual is named as a “position” in FPO. Only the competitive, powerful positions selected as elites could achieve the limited opportunity to update. The elite team reduces the possibility of all of the individuals moving toward the same place. Eight well-known benchmark functions are used to test the performance of FPO. Four typical multimodal benchmark functions are used to test the global search ability of FPO and four fixed-dimension multimodal optimization problems are selected to make a comparison of the convergence rate between several well-known algorithms. The experimental results show that the FPO algorithm is able to provide excellent exploitation, utilizing local minima avoidance and exploration simultaneously.

We report on fault-tolerant technology for use with high-speed parallel evolutionary computation on many-core processors. In particular, for distributed GA models which communicate between islands, we propose a method where an island's ID number is added to the header of data transferred by this island for use in fault detection, and we evaluate this method using Deceptive functions and Sudoku puzzles. As a result, we show that it is possible to detect single stuck-at faults with practically negligible overheads in applications where the time spent performing genetic operations is large compared with the data transfer speed between islands. We also show that it is still possible to obtain an optimal solution when a single stuck-at fault is assumed to have occurred, and that increasing the number of parallel threads has the effect of making the system less susceptible to faults and more sustainable.

Human tracking in monocular image sequences has been studied in the field of computer vision for many kinds of applications such as surveillance system, intelligent room, sports video analysis and so on. Human tracking in real environment is challenging topic due to various factors such as illumination change, partial or almost complete occlusion of human body, and wide variety of body shapes. In this paper, we present a robust human tracking using statistical human shape model of appearance variation with postural change. Our part-based statistical human model can generate learned appearances of main human poses, and enables effective and robust human tracking with simple features such silhouette, edge and color. Our proposed method achieves human tracking robust not only to partial occlusion but also to postural change. The experimental results validate the robustness of our methods in the real indoor environments.

In this paper, we use the problem of solving Sudoku puzzles to demonstrate the possibility of achieving practical processing time through the use of GPUs for parallel processing in the application of genetic computation to problems for which the use of genetic computing has not been investigated before because of the processing time problem. To increase accuracy, we propose a genetic operation that takes building-block linkage into account. As a parallel processing model for higher performance, we use a multiple-population coarse-grained GA model to counter initial value dependence under the condition of a limited number of individuals. Specifically, we show that it is possible to reach a solution in a few seconds of processing time with a correct solution rate of 100%, even for extremely difficult problems by parallel processing of genetic computation on a GeForce GTX 460, a commercial GPU produced by the NVIDIA Corporation.

The use of low-cost multi-core processors that target small-scale parallelism of several or several tens of processing units has recently spread to general-purpose personal computers. As part of this trend, research on converting evolutionary computation to parallel processing on multi-core processors has begun. In our research, we place importance on proposing multi-core processor architectures from the viewpoint of evolutionary computation and on proposing indices for identifying chips that exhibit high affinity with evolutionary computation from among existing multi-core processors. This paper takes a first step in this direction by focusing on local memory built into a multi-core processor and proposing a specification that allows the contents of local memory in any core to be directly accessed by any other core without using main memory. We describe a desktop comparison with single-core processors and general multi-core processors and show that a speed-up effect should be expected with the proposed specification for both fine-grain and coarse-grained parallelization techniques. We also describe an evaluation on actual multi-core processor for the case of fine-grain parallelization and show that the speed-up effect is also useful in reducing energy consumption.

This paper describes proposing multi-core processor architectures from the viewpoint of evolutionary computation. We focus on local memory built into a multi-core processor and propose a specification that allows the contents of local memory in any core to be directly accessed by any other core without using main memory. We describe an evaluation on an actual multi-core processor and show that the speed-up effect and reducing energy consumption should be expected with the proposed specification.

In this paper we report on the effect of viral infection with tropism on the formation of building blocks in genetic operations. In previous research, we applied genetic algorithms to the analysis of time-series signals with noise. We demonstrated the possibility of reducing the number of required entities and improving the rate of convergence when searching for a solution by having some of the host chromosomes harbor viruses with a tropism function. Here, we simulate problems having both multimodality and deceptiveness features and problems that include noise as test functions, and show that viral infection with tropism can increase the proportion of building blocks in the population when it cannot be assumed that a necessary and sufficient number of entities are available to find a solution. We show that this capability is especially noticeable in problems that include noise.

In this paper, we discuss dependencies on player formation when using a classifier system in a decision algorithm for agents in a soccer game. Our aim is to respond to the changing environment of video gaming that has resulted from the growth of the Internet, and to provide bug-free programs in a short time. We have already proposed a bucket brigade algorithm and a procedure for choosing what to learn depending on the frequency of events with the aim of facilitating real-time learning while a game is in progress. We have also proposed a hybrid system configuration that combines existing algorithm strategies with a classifier system, and we have reported on the effectiveness of this hybrid system. In this paper, we pit players in several different formations against each other and show that the proposed system is able to learn regardless of the differences in formation. We also show that by performing simulations ahead of time, it is possible to investigate formations that will be effective against an opponent’s formation. Finally, by investigating changes in frequency and success rates for each type of play due to changes in formation, we show that it is possible to acquire a team strategy for the current formation through learning.

In this paper we report on the effect of viral infection with tropism on the formation of building blocks in genetic operations. In previous research, we applied genetic algorithms to the analysis of time-series signals with noise. We demonstrated the possibility of reducing the number of required entities and improving the rate of convergence when searching for a solution by having some of the host chromosomes harbor viruses with a tropism function. Here, we simulate problems having both multimodality and deceptiveness features and problems that include noise as test functions, and show that viral infection with tropism can increase the proportion of building blocks in the population when it cannot be assumed that a necessary and sufficient number of entities are available to find a solution. We show that this capability is especially noticeable in problems that include noise.

In this paper, we discuss dependencies on player formation when using a classifier system in a decision algorithm for agents in a soccer game. Our aim is to respond to the changing environment of video gaming that has resulted from the growth of the Internet, and to provide bug-free programs in a short time. We have already proposed a bucket brigade algorithm and a procedure for choosing what to learn depending on the frequency of events with the aim of facilitating real-time learning while a game is in progress. We have also proposed a hybrid system configuration that combines existing algorithm strategies with a classifier system, and we have reported on the effectiveness of this hybrid system. In this paper, we pit players in several different formations against each other and show that the proposed system is able to learn regardless of the differences in formation. We also show that by performing simulations ahead of time, it is possible to investigate formations that will be effective against an opponent's formation. Finally, by investigating changes in frequency and success rates for each type of play due to changes in formation, we show that it is possible to acquire a team strategy for the current formation through learning.

This paper describes our study into the concept of using rewards in a classifier system applied to the acquisition of decision-making algorithms for agents in a soccer game. Our aim is to respond to the changing environment of video gaming that has resulted from the growth of the Internet, and to provide bug-free programs in a short time. We have already proposed a bucket brigade algorithm and a procedure for choosing what to learn depending on the frequency of events with the aim of facilitating real-time learning while a game is in progress. We have also proposed a hybrid system configuration that combines existing algorithm strategies with a classifier system, and we have reported on the effectiveness of this hybrid system. This paper proposes applying genetic algorithms to the search for rewards in reinforcement learning where designers have hitherto used empirical trial-and-error methods. By pitting this new technique against an existing soccer game with algorithms designed by humans, we demonstrate the possibility of using genetic algorithms to automate the setting of rewards from the viewpoint of achieving a greater success rate and faster convergence than in cases where the success rewards for each play are set by the designers based on trial and error.

In this paper we report on a study in which genetic algorithms are applied to the analysis of noisy time-series signals, which is related to the problem of analyzing the motion characteristics of moving bodies (distance, bearing, course, velocity, etc.) by covertly sampling the sound of moving objects with submarine monitoring systems that track moving objects travelling on or through the water. In particular, we propose improving the system's ability to search through noisy data by grafting viruses onto the chromosomes used in genetic algorithms. Specifically, we propose a search method that can cope robustly with noise through the cooperative action of a wide-area search implemented by host chromosomes and a local search implemented by viruses grafted onto these chromosomes. To improve the infection rate, we also impose limits on the types of host entity that can be infected by viruses. By conducting evaluation tests in computer simulations, we show that the proposed technique can achieve a better rate of convergence and is capable of searching for a solution with fewer entities.

This paper describes our study into the concept of using rewards in a classifier system applied to the acquisition of decision-making algorithms for agents in a soccer game. Our aim is to respond to the changing environment of video gaming that has resulted from the growth of the Internet, and to provide bug-free programs in a short time. We have already proposed a bucket brigade algorithm (a reinforcement learning method for classifiers) and a procedure for choosing what to learn depending on the frequency of events with the aim of facilitating real-time learning while a game is in progress. We have also proposed a hybrid system configuration that combines existing algorithm strategies with a classifier system, and we have reported on the effectiveness of this hybrid system. In this paper, we report on the results of performing reinforcement learning with different reward values assigned to reflect differences in the roles performed by forward, midfielder and defense players, and we describe the results obtained when learning is performed with different combinations of success rewards for various type of play such as dribbling and passing. In 200 matches played against an existing soccer game incorporating an algorithm devised by humans, a better win ratio and better convergence were observed compared with the case where learning was performed with no roles assigned to all of the in-game agents.

We have researched for the applicability of the genetic algorithms (GA) to problems such as multi objective optimization, time series prediction, the analysis from observed noisy data and the solution of implicit functions. Concerning these problems, we reported that GA is effective for the tracking of the moving ships and for the tracking of the objects orbiting the earth from observed time series bearing and distance data or only from sound signals radiated by of unknown moving object. In this paper, we report on the analysis for the movement characteristics (distance, bearing, velocity, course) and sound source frequency by elimination Doppler effect from sounds radiated by the propulsion system of an unknown object in sea. To analyze those data, we applied two kinds of GA. One of them analyzes harmonic frequencies and their phases of the Doppler effected sound signals. The other GA analyzes the movement characteristics and sound source frequency based on the output of the former GA. We could prove the applicability of GA to this analysis through a computer simulation.

We have researched for the applicability of the genetic algorithms (GA) to issues such as multi objective optimization, time series prediction, the analysis from observed noisy data and the solution of implicit functions. Concerning these problems, we reported that GA is effective for the tracking of the moving ships, and for the tracking of the objects orbiting the earth from observed time series bearing and distance data. In this paper, we report on the analysis for the movement characteristics of an object in sea water from an observed time series sound signal. This analysis is more complex than those of earlier reports. To analyze the movement characteristics, we applied a two-step GA analysis. The first step GA analyzes harmonic frequencies and their phases of the sound signal radiated by the moving object. The second step GA analyzes the movement characteristics by using output of the first step GA. We could prove the applicability of GA to this analysis through a computer simulation.

In this paper, a report is released in relation to an example of the application of the genetic algorithm based on a local evaluation rule where attention is paid to a region in a communication circle from an access point taking up an access point layout problem in the wireless LAN construction as an objective. When the optimum solution is probed by means of an evaluation function taking up the whole entity of the LAN, a circuit becomes unstable in a specific service area even if a solution is the quasi-optimum solution as the whole entity of the system in case that attention is paid to the individual clients, there might be a possibility that an area not necessarily successful in realizing high-quality service is in existence. In this paper, it is explained that the method hereby proposed can first of all obtain the quasi-optimum solution taking up the whole entity of the wireless LAN system, as with the case that an evaluation function taking up the whole entity of the system as an objective is applied. At the next stage, it is explained whether there is a possibility that a solution for practical use where clients' communication situation is considered in a service area will be obtained.

This paper reports on the application of classifier systems to the acquisition of decision-making algorithms for agents in online soccer games. The objective of this research is to support changes in the videogame environment brought on by the Internet and to enable the provision of bug-free programs in a short period of time. To achieve real-time learning during a game, a bucket brigade algorithm is used to reinforce learning by classifiers and a technique for selecting learning targets according to event frequency is adopted. A hybrid system combining an existing strategy algorithm and a classifier system is also employed. In experiments that observed the outcome of 10,000 soccer games between this event-driven classifier system and a human-designed algorithm, the proposed system was found to be capable of learning effective decision-making algorithms in real time.

This paper reports on the application of classifier systems to the acquisition of decision-making algorithms for agents in online soccer games. The objective of this research is to support changes in the video-game environment brought on by the Internet and to enable the provision of bug-free programs in a short period of time. To achieve real-time learning during a game, a bucket brigade algorithm is used to reinforce learning by classifiers and a technique for selecting learning targets according to event frequency is adopted. A hybrid system combining an existing strategy algorithm and a classifier system is also employed. In experiments that observed the outcome of 10,000 soccer games between this event-driven classifier system and a human-designed algorithm, the proposed system was found to be capable of learning effective decision-making algorithms in real time.

Unlike conventional haptic devices, an encountered-type device is not held by a user all the time. Instead, the device stays at the location of a virtual object and waits for the user to encounter it. In this paper, we extend this concept to fingertip contacts and design an encountered-type haptic display for multiple fingertip contacts to simulate tasks of grasping an object with any shape and size. Before designing the device, we intensively observed human grasping behaviors. This observation was very helpful to determine the mechanism of the device. An encountered-type device for three-fingered grasping was actually prototyped based on our design. We then discuss a path planning algorithm for the designed device, once again based on the observation of human grasping behaviors. A rough sketch of the algorithm is proposed and a preliminary experimental result to support our idea is shown.

In a previous report, we demonstrated the effectiveness of genetic algorithms in tracking a moving object with a destination as it attempts to camouflage its movement to avoid pursuit or attack. In this paper, we consider more complex object motion and show that a system that applies genetic algorithms can obtain solutions with good accuracy compared to the conventional system.

A new idea for evolvable hardware based on a microprocessor is proposed. Evolvable hardware is a new direction in hardware research that fuses evolutionary computation and reconfigurable logic LSI circuits. In recent years, there has been much research using programmable logic devices (PLDs) and field programmable gate arrays (FPGAs). In particular, the application of digital circuit evolution to engineering fields has already begun. On the other hand, the long learning time, the difficulty of predicting when an effective capability will appear, the large chip size and other such problems have hindered progress in diffusion into engineering fields. We propose RTL (register transfer level) evolution performed on a microprocessor as a means of addressing these problems, Specifically, we propose: (1) incorporating flash memory into the microprocessor to allow on-board programming and reprogramming, (2) using genetic algorithms to provide an RTL learning capability, and (3) the use of a framework that provides for the coexistence of static programs and programs that self-organize through learning. On the basis of a simple hand-design, we concluded that the proposed method is more effective in terms of learning efficiency and reliability than the conventional approach using FPGAs and PLDs.

Presents new principles for online monitoring in the context of multiprocessors (especially massively parallel processors) and then focuses on the effect of the aliasing probability on the error detection process. In the proposed test architecture, concurrent testing (or online monitoring) at the system level is accomplished by enforcing the run-time testing of the data and control dependences of the algorithm currently being executed on the parallel computer. In order to help in this process, each message contains both source and destination addresses. At each message source, the sequence of destination addresses of the outgoing messages is compressed on a block basis. At the same time, at each destination, the sequence of source addresses of all incoming messages is compressed, also on a block basis. Concurrent compression of the instructions executed by the PEs is also possible. As a result of this procedure, an image of the data dependences and of the control flow of the currently running algorithm is created. This image is compared, at the end of each computational block, with a reference image created at compilation time. The main results of this work are in proposing new principles for the online system-level testing of multiprocessor systems, based on signaturing and monitoring the data dependences together with the control dependences, and in providing an analytical model and analysis for the address compression process used for monitoring the data routing process.>

Presents new principles for online monitoring in the context of multiprocessors (especially massively parallel processors) and then focuses on the effect of the aliasing probability on the error detection process. In the proposed test architecture, concurrent testing (or online monitoring) at the system level is accomplished by enforcing the run-time testing of the data and control dependences of the algorithm currently being executed on the parallel computer. In order to help in this process, each message contains both source and destination addresses. At each message source, the sequence of destination addresses of the outgoing messages is compressed on a block basis. At the same time, at each destination, the sequence of source addresses of all incoming messages is compressed, also on a block basis. Concurrent compression of the instructions executed by the PEs is also possible. As a result of this procedure, an image of the data dependences and of the control flow of the currently running algorithm is created. This image is compared, at the end of each computational block, with a reference image created at compilation time. The main results of this work are in proposing new principles for the online system-level testing of multiprocessor systems, based on signaturing and monitoring the data dependences together with the control dependences, and in providing an analytical model and analysis for the address compression process used for monitoring the data routing process. >