tailieunhanh - Báo cáo khoa học: "hệ cộng dồn "mùi" cải tiến tron tối ưu hóa bầy kiến"

Tuyển tập báo cáo nghiên cứu khoa học của trường đại học nông nghiệp 1 đề tài: hệ cộng dồn "mùi" cải tiến tron tối ưu hóa bầy kiến. | Tạp chí KHKT Nông nghiệp 2007 Tập V Sô 4 60-66 ĐẠI HỌC NÔNG NGHIỆP I HỆ CỌNG DON MÙI CAI TIẾN TRONG TÔI ƯU HÓA BẦY KIẾN An Improved Aggregation Pheromone System in the Ant Colony Optimization Nguyễn Hoàng Huy Nguyễn Hải Thanh SUMMARY As a bio-inspired computational paradigm Ant colony optimization ACO has been applied with great success to a large number of discrete optimization problems. However up to now there are few adaptations of ACO to continuous optimization problems whereas these problems are frequent occurrence. Moreover almost all of the adaptations use marginal distribution models and the pheromone update rules used are quite different than those of the original ACO algorithms. In some recent papers Shigeyoshi Tsutsui and colleagues have proposed two algorithms for continuous optimization called the Aggregation Pheromone System APS and the enhanced APS eAPS . These algorithms apply the same pheromone update rule in a way similar to those of the original ACO algorithms and as a result the aggregation pheromone density eventually becomes a mixture of multivariate normal probability density functions. However both of the above algorithms do not guarantee to find out a solution converging to an optimal solution. Based on an insight into the mathematical techniques used to prove convergence of ACO algorithms on the discrete domain we propose an improved APS iAPS . iAPS inherits APS s ant-colony based approaches and allows a stronger exploration of better solutions found and at the same time it can prevent premature stagnation of the search. Consequently iAPS has a higher probability of finding out an optimal solution. We hope iAPS will be applied for realistic optimization problems in agricultural fields. Keywords Aggregation pheromone system Ant colony optimization ACO approximation algorithm metaheuristics. 1. ĐẶT VẤN ĐỀ Trong các nghiên cứu về nông nghiệp và sinh học chúng ta gặp nhiều bài toán tối ưu. Nhiệm vụ chính của các bài toán này là phải xây dựng .