• Journal of Computing Science and Engineering
• /
• v.8 no.4
• /
• pp.199-206
• /
• 2014

The ant colony optimization (ACO) algorithm is a new heuristic algorithm that offers good robustness and searching ability. With in-depth exploration, the ACO algorithm exhibits slow convergence speed, and yields local optimization solutions. Based on analysis of the ACO algorithm and the genetic algorithm, we propose a novel hybrid genetic ant colony optimization (NHGAO) algorithm that integrates multi-population strategy, collaborative strategy, genetic strategy, and ant colony strategy, to avoid the premature phenomenon, dynamically balance the global search ability and local search ability, and accelerate the convergence speed. We select the traveling salesman problem to demonstrate the validity and feasibility of the NHGAO algorithm for solving complex optimization problems. The simulation experiment results show that the proposed NHGAO algorithm can obtain the global optimal solution, achieve self-adaptive control parameters, and avoid the phenomena of stagnation and prematurity.

• Journal of Information Processing Systems
• /
• v.14 no.2
• /
• pp.448-454
• /
• 2018

Artificial bee colony (ABC) algorithm has attracted significant interests recently for solving the multivariate optimization problem. However, it still faces insufficiency of slow convergence speed and poor local search ability. Therefore, in this paper, a modified ABC algorithm with bees' number reallocation and new search equation is proposed to tackle this drawback. In particular, to enhance solution accuracy, more bees in the population are assigned to execute local searches around food sources. Moreover, elite vectors are adopted to guide the bees, with which the algorithm could converge to the potential global optimal position rapidly. A series of classical benchmark functions for frequency-modulated sound waves are adopted to validate the performance of the modified ABC algorithm. Experimental results are provided to show the significant performance improvement of our proposed algorithm over the traditional version.

• Korean Management Science Review
• /
• v.17 no.2
• /
• pp.39-54
• /
• 2000

This paper presents a causal forecasting model using guided genetic algorithm in continuous manufacturing process. The guide genetic algorithm(GGA) is an extended genetic algorithm(GA) using penalty function and population diversity index to increase forecasting accuracy. GGA adds to the canonical GA the concept of a penalty function to avoid selecting the unproductive chromosomes and to make a proper searching direction. Also, GGA modifies the current population using the similarity of chromosomes to avoid falling into the trap of local optimal solution. For investigation GGA performance, we used a set of real data that was collected in local glass melting processes, and experimental results show the proposed model results in the better forecasting accuracy than linear regression model and canonical GA.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.5 no.4
• /
• pp.281-285
• /
• 2005

Learning and evolving are two basics for data mining. As compared with classical learning theory based on objective function with minimizing training errors, the recently evolutionary computing has had an efficient approach for constructing optimal model without the minimizing training errors. The global search of evolutionary computing in solution space can settle the local optima problems of learning models. In this research, combining co-evolving algorithm into statistical learning theory, we propose an co-evolutionary computing for statistical learning theory for overcoming local optima problems of statistical learning theory. We apply proposed model to classification and prediction problems of the learning. In the experimental results, we verify the improved performance of our model using the data sets from UCI machine learning repository and KDD Cup 2000.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.5
• /
• pp.1303-1311
• /
• 1999

In order to learn in dynamic environments, reinforcement learning algorithms like Q-learning, TD(0)-learning, TD(λ)-learning have been proposed. however, most of them have a drawback of very slow learning because the reinforcement value is given when they reach their goal state. In this thesis, we have proposed a reinforcement learning method that can approximate fast to the goal state in maze environments. The proposed reinforcement learning method is separated into global learning and local learning, and then it executes learning. Global learning is a learning that uses the replacing eligibility trace method to search the goal state. In local learning, it propagates the goal state value that has been searched through global learning to neighboring sates, and then searches goal state in neighboring states. we can show through experiments that the reinforcement learning method proposed in this thesis can find out an optimal solution faster than other reinforcement learning methods like Q-learning, TD(o)learning and TD(λ)-learning.

• ETRI Journal
• /
• v.44 no.2
• /
• pp.313-326
• /
• 2022

This study presents a noble scheme for distributed and parallel simulations with optimized agent placement for simulation instances. The traditional parallel simulation has some limitations in that it does not provide sufficient performance even though using multiple resources. The main reason for this discrepancy is that supporting parallelism inevitably requires additional costs in addition to the base simulation cost. We present a comprehensive study of parallel simulation architectures, execution flows, and characteristics. Then, we identify critical challenges for optimizing large simulations for parallel instances. Based on our cost-benefit analysis, we propose a novel approach to overcome the performance constraints of agent-based parallel simulations. We also propose a solution for eliminating the synchronizing cost among local instances. Our method ensures balanced performance through optimal deployment of agents to local instances and an adaptive agent placement scheme according to the simulation load. Additionally, our empirical evaluation reveals that the proposed model achieves better performance than conventional methods under several conditions.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.32 no.1
• /
• pp.72-78
• /
• 2009

This paper considers a multiagent scheduling problem under public information where a machine is shared by multiple agents. Each agent has a local objective among the minimization of total completion time and the minimization of maximum. In this problem, it is assumed that scheduling information is public. Therefore an agent can access to complete information of other agents and pursue efficient schedules in a centralized manner. We propose an enumeration scheme to find Pareto optimal schedules and a multiobjective genetic algorithm as a heuristic approach. Experimental results indicate that the proposed genetic algorithm yields close-to Pareto optimal solution under a variety of experimental conditions.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.27 no.4
• /
• pp.1-10
• /
• 2002

The ring routing and wavelength assignment problem arose in the planning of optical communication networks which use WDM rings. Traffic demands are given for each pair of nodes in an ring : each demand must be routed one of the two possible connections round the ring and the wavelength assignments must be made so that there are no conflicts : that is. no two connections whose routes share a link can be assigned the same wavelength along that link. The objective is to minimize the number of used wavelengths. We propose the local optimal routing for the problem and show that there always exists an optimal solution satisfying it. Furthermore we suggest a new lower bound for the problem and show that it is very efficient for the worst case example.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.174-174
• /
• 2000

Q-learning, based on discrete state and action space, is a most widely used reinforcement Learning. However, this requires a lot of memory and much time for learning all actions of each state when it is applied to a real mobile robot navigation using continuous state and action space Region-based Q-learning is a reinforcement learning method that estimates action values of real state by using triangular-type action distribution model and relationship with its neighboring state which was defined and learned before. This paper proposes a new Region-based Q-learning which uses a reward assigned only when the agent reached the target, and get out of the Local optimal path with adjustment of random action rate. If this is applied to mobile robot navigation, less memory can be used and robot can move smoothly, and optimal solution can be learned fast. To show the validity of our method, computer simulations are illusrated.

• Journal of the military operations research society of Korea
• /
• v.30 no.2
• /
• pp.81-95
• /
• 2004

This paper presents optimal solutions for the following two problems ; (1) The vehicle routing problem with deadline considering homeward-route transport requirement, (2) The problem of (1) with relaxed constraint that the vehicle cannot visit the depot during the trip. That is, for the second problem, vehicle can visit the depot several times during the trip. The formulation and optimal solution we obtained outperform the existing ones proposed by other researchers in the sense of time and total distance traveled and the results can be applicable to short-haul local transport.