Improved Simulated Annealing With Cobweb Spreading In Continuous Space
Mahdi Imani, Seyede Fatemeh Ghoreishi, Masoud Shariat-Panahi · 2014
ABSTRACT: An innovative approach based on Simulated Annealing method to optimize the problems in continuous space is introduced in this paper. The Simulated Annealing is a popular method which finds the optimum value of functions based on temperature changes during its search. Specifying the trend of temperature change and step length is the most critical issue in the original SA method. The SA algorithm with large or small initial step length cannot reach the optimum value of the function efficiently. The initial temperature and its trend of decrease affect the results of the search in the space. In addition, swarm algorithms like PSO and GA are the powerful methods in optimization. In this paper, a new method called Cobweb Simulated Annealing (CSA) with swarm search in the continuous space is presented. The number of population, temperature and step length are adaptive during the search in this algorithm. The searching points spread to explore the entire search space especially in the first stages. This method is applied to several benchmark functions and the results have shown its reliability and efficiency to find the optimum value of functions in comparison with some powerful modifications of SA. CSA is able to search more extensive area of the whole space with less computational cost. The other significant capability of CSA algorithm is finding more local minimums than other modified algorithms.