Toshiharu Hatanaka

Improvement of Particle Swarm Optimization for High-Dimensional Space

Particle swarm optimization (PSO) is a population-based search methodology inspired by social behavior observed in nature, such as flocks of birds and schools of fish. In many studies, PSO has been successful in a variety of optimization problems. The purpose of this paper is to improve performance of the PSO algorithm in case of high-dimensional problems. We propose a novel PSO model, the rotated particle swarm (RPS), which is introduced the coordinate conversion. The numerical simulation results show the RPS is effective in optimizing high-dimensional functions

Evolution Strategies Based Particle Filters for Nonlinear State Estimation

There has been significant recent interest of particle filters for nonlinear state estimation. Particle filters evaluate a posterior probability distribution of the state variable based on observations in Monte Carlo simulation using so-called importance sampling. However, degeneracy phenomena in the importance weights deteriorate the filter performance. By recognizing the similarities and the difference of the processes between the particle filters and Evolution Strategies, a new filter, Evolution Strategies Based Particle Filter, is proposed to circumvent this difficulty and to improve the performance. The applicability of the proposed idea is illustrated by numerical studies.

Evolution strategies based particle filters for state and parameter estimation on nonlinear models

The massive increase of computational power has led to the rebirth of Monte Carlo integration and its application of Bayesian filtering, or particle filters. Particle filters evaluate a posterior probability distribution of the state variable based on observations in Monte Carlo simulation using so-called importance sampling. However, the filter performance is degraded by degeneracy phenomena in the importance weights. A filter called the evolution strategies (ES)-based particle filter has been proposed to circumvent this difficulty and improve the performance by recognizing the similarities and the difference between the particle filters and ES. The SIE filter is applied to simultaneous state and parameter estimation of nonlinear state space models. Numerical simulation studies have been conducted to exemplify the applicability of this approach.

Eggplant classification using artificial neural network

Recently there have been developed automatic grading and sorting systems for fruits and vegetables. In this paper, eggplant grading system using image processing and artificial neural network is considered. The lighting conditions are discussed for taking color components of the eggplant image effectively. The shape parameters such as length, girth, etc. are measured using image processing. On the other hand, bruises of the eggplants are detected and classified based on the color information by using artificial neural network. Some experimental results are presented for illustration.

Block oriented nonlinear model identification by evolutionary computation approach

Block oriented nonlinear models such as Hammerstein and Wiener models are composed by a cascade combination of a linear dynamic model and a static or memoryless nonlinear function. Though these have very simple model structures, they can represent and approximate many real processes especially in electrical, chemical and biological engineering. In this paper, a novel approach for nonlinear system identification is addressed for the block oriented models. After approximating the nonlinear static part or its inverse by a piecewise linear function, its parameters are estimated by using the evolutionary computation approach such as genetic algorithm (GA) and evolution strategies (ES), while the linear dynamic system part is estimated by the least squares method. Numerical simulation studies illustrate the applicability of the proposed approach.

Multi-objective structure selection for radial basis function networks based on genetic algorithm

Radial basis function (RBF) network is well known as a good performance approach to nonlinear system modeling. Though structure selection of RBF network is an important issue, the framework of this problem has not been established. In this paper, we propose multiobjective structure selection method for RBF networks based on MOGA (multiobjective genetic algorithm). The structure of RBF networks is encoded to the chromosomes in GA, then evolved toward to Pareto-optimum for multiobjective functions concerned with model accuracy and complexity. Some numerical simulation results indicate the applicability of the proposed approach.

Training Hidden Markov Model Structure with Genetic Algorithm for Human Motion Pattern Classification

Physical exercise classification method by hidden Markov model (HMM) is considered in this study. The aim of this study is to discuss the availability of HMM based motion modeling in order to compare human skills. In this paper, a preprocessing technique for observed human motion by self-organizing map (SOM) to label a motion characteristic is proposed. Then, HMM construction method by using genetic algorithm (GA) with Baum-Welch algorithm, modified crossover and mutation is introduced. Simulation studies are carried out for bat swing motions. It is shown that the proposed approach has an ability to recognize bat swing motions

Pattern Classification by Evolutionary RBF Networks Ensemble Based on Multi-objective Optimization

In this paper, evolutionary multi-objective selection method of RBF networks structure and its application to the ensemble learning is considered. The candidates of RBF network structure are encoded into the chromosomes in GAs. Then, they evolve toward Pareto-optimal front defined by several objective functions concerning with model accuracy, model complexity and model smoothness. RBF network ensemble is constructed of the obtained Pareto-optimal models since such models are diverse. This method is applied to the pattern classification problem. Experiments on the benchmark problem demonstrate that the proposed method has comparable generalization ability to conventional ensemble methods.

System parameter estimation by evolutionary strategy

This paper discusses an application of Evolutionary Strategy (ES), a simple direct probabilistic search and optimization algorithm to system parameter estimation. We consider multiple estimates of the system parameters, which interactively or independently seeks for the better estimates to fit the environment, and construct an estimate by some algebraic operations on these estimates. Illustrating numerical examples show the applicability of the proposed idea from the viewpoints of adaptability and robustness.

Nonlinear Dynamic System Identification Based on Multiobjectively Selected RBF Networks

In this paper, nonlinear dynamic system identification by using multiobjectively selected RBF network is considered. RBF networks are widely used as a model structure for nonlinear systems. The determination of its structure that is the number of basis functions is prior important step in system identification, and the tradeoff between model complexity and accuracy exists in this problem. By using multiobjective evolutionary algorithms, the candidates of the RBF network structure are obtained in the sense of Pareto optimality. We discuss an application to system identification by using such RBF networks having Pareto optimal structures. Some numerical simulations for nonlinear dynamic systems are carried out to show the applicability of the proposed approach.