Thi Thuy Ngo

A cooperative particle swarm optimizer with stochastic movements for computationally expensive numerical optimization problems

Abstract Nature is the rich principal source for developing optimization algorithms. Metaheuristic algorithms can be classified with the emphasis on the source of inspiration into several categories such as biology, physics, and chemistry. The particle swarm optimization (PSO) is one of the most well-known bio-inspired optimization algorithms which mimics movement behavior of animal flocks especially bird and fish flocking. In standard PSO, velocity of each particle is influenced by the best individual and its best personal experience. This approach could make particles trap into the local optimums and miss opportunities of jumping to far better optimums than the currents and sometimes causes fast premature convergence. To overcome this issue, a new movement concept, so called extraordinariness particle swarm optimizer (EPSO) is proposed in this paper. The main contribution of this study is proposing extraordinary motion for particles in the PSO. Indeed, unlike predefined movement used in the PSO, particles in the EPSO can move toward a target which can be global best, local bests, or even the worst individual. The proposed improved PSO outperforms than the standard PSO and its variants for benchmarks such as CEC 2015 benchmarks. In addition, several constrained and engineering design problems have been tackled using the improved PSO and the optimization results have been compared with the standard PSO, variants of PSO, and other optimizers.

KU Battle of Metaheuristic Optimization Algorithms 1: Development of Six New/Improved Algorithms

Each of six members of hydrosystem laboratory in Korea University (KU) invented either a new metaheuristic optimization algorithm or an improved version of some optimization methods as a class project for the fall semester 2014. The objective of the project was to help students understand the characteristics of metaheuristic optimization algorithms and invent an algorithm themselves focusing those regarding convergence, diversification, and intensification. Six newly developed/improved metaheuristic algorithms are Cancer Treatment Algorithm (CTA), Extraordinary Particle Swarm Optimization (EPSO), Improved Cluster HS (ICHS), Multi-Layered HS (MLHS), Sheep Shepherding Algorithm (SSA), and Vision Correction Algorithm (VCA). This paper describes the details of the six developed/improved algorithms. In a follow-up companion paper, the six algorithms are demonstrated and compared through well-known benchmark functions and a real-life engineering problem.

Comparison of Parameter-Setting-Free and Self-adaptive Harmony Search

This study compares the performance of all parameter-setting-free and self-adaptive harmony search algorithms proposed in the previous studies, which do not ask for the user to set the algorithm parameter values. Those algorithms are parameter-setting-free harmony search, Almost-parameter-free harmony search, novel self-adaptive harmony search, self-adaptive global-based harmony search algorithm, parameter adaptive harmony search, and adaptive harmony search, each of which has a distinctively different mechanism to adaptively control the parameters over iterations. Conventional mathematical benchmark problems of various dimensions and characteristics and water distribution network design problems are used for the comparison. The best, worst, and average values of final solutions are used as performance indices. Computational results show that the performance of each algorithm has a different performance indicator depending on the characteristics of optimization problems such as search space size. Conclusions derived in this study are expected to be beneficial to future research works on the development of a new optimization algorithm with adaptive parameter control. It can be considered to improve the algorithm performance based on the problem’s characteristic in a much simpler way.

The Extraordinary Particle Swarm Optimization and Its Application in Constrained Engineering Problems

The particle swarm optimization (PSO) is a natural-inspire optimization algorithm mimicking the movement behavior of animal flocks for food searching. Although the algorithm presents some advantages and widely application, however, there are several drawbacks such as trapping in local optima and immature convergence rate. To overcome these disadvantages, many improved versions of PSO have been proposed. One of the latest variants is the extraordinary particle swarm optimization (EPSO). The particles in the EPSO are assigned to move toward their own determined target through the search space. The applicability of EPSO is verified by several experiments in engineering optimization problems. The application results show the outperformance of the EPSO than the other PSO variants in terms of solution searching and as well as convergence rate.

Numerical Solution of Boundary Value Problems Using Artificial Neural Networks and Harmony Search

In this paper, we present an algorithm based on artificial neural networks (ANNs) and harmony search (HS) for the numerical solution of boundary value problems (BVPs), which evolves in most of the science and engineering applications. An approximate trial solution of the BVPs is constructed in terms of ANN in a way that it satisfies the desired boundary conditions of the differential equation (DE) a utomatically. Approximate satisfaction of the trial solution results in an unsupervised error, which is minimized by training ANN using the harmony search algorithm (HSA). A BVP modeling the flow of a stretching surface is considered here as a test problem to validate the accuracy, convergence and effectiveness of the proposed algorithm. The obtained results are compared with the available exact solution also to test the correctness of the algorithm.

A New Collaborative Approach to Particle Swarm Optimization for Global Optimization

Particle swarm optimization (PSO) is population-based metaheuristic algorithm which mimics animal flocking behavior for food searching and widely applied in various fields. In standard PSO, movement behavior of particles is forced by the current bests, global best and personal best. Despite moving toward the current bests enhances convergence, however, there is a high chance for trapping in local optima. To overcome this local trapping, a new updating equation proposed for particles so-called extraordinary particle swarm optimization (EPSO). The particles in EPSO move toward their own targets selected at each iteration. The targets can be the global best, local bests, or even the worst particle. This approach can make particles jump from local optima. The performance of EPSO has been carried out for unconstrained benchmarks and compared to various optimizers in the literature. The optimization results obtained by the EPSO surpass those of standard PSO and its variants for most of benchmark problems.

Linear system theory-based optimization of detention basin's location and size at watershed scale

AbstractIn the context of urbanization and waterworks construction, detention basins located in upstream tributaries have become one of the most-effective technical solutions for downstream flood reduction. This technical note proposes a new optimal approach for determining the most-effective design combination of detention basins (in terms of their sizes and locations) in a watershed utilizing linear reservoir theory. The optimal detention basin is obtained by minimizing the deviation of downstream peak discharge from river capacity discharge. With the presence of a detention basin in an upstream river, the flood peak downstream decreases by 33%, corresponding to two scenarios of rainfall duration with a return period T=100 years. The storage volume estimation of the optimal detention reservoir proposed in the present paper demonstrates the high applicability of this optimization approach.

KU Battle of Metaheuristic Optimization Algorithms 2: Performance Test

In the previous companion paper, six new/improved metaheuristic optimization algorithms developed by members of Hydrosystem laboratory in Korea University (KU) are introduced. The six algorithms are Cancer Treatment Algorithm (CTA), Extraordinary Particle Swarm Optimization (EPSO), Improved Cluster HS (ICHS), Multi-Layered HS (MLHS), Sheep Shepherding Algorithm (SSA), and Vision Correction Algorithm (VCA). The six algorithms are tested and compared through six well-known unconstrained benchmark functions and a pipe sizing problem of water distribution network. Performance measures such as mean, best, and worst solutions (under given maximum number of function evaluations) are used for the comparison. Optimization results are obtained from thirty independent optimization trials. Obtained Results show that some of the newly developed/improved algorithms show superior performance with respect to mean, best, and worst solutions when compared to other existing algorithms.