8642

8023

كارشناسي ارشد

هوش مصنوعي

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1392

نه،64ص.: مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

عبدالرضا ميرزايي

محمدعلي منتظري

16/1/93

مهران صفاياني، ناصر قديري مدرس

مهندسي برق و كامپيوتر

ID8023

به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي

Improving Evolutionary Algorithms with Machine Learning Mohammadtaghi Moein m moien@ec iut ac ir Data of Submission 22 09 2013 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language Farsi Supervisor Abdolreza Mirzaei mirzaei@cc iut ac ir Abstract Machine Learning is one the most promising and salient research area in artificial intelligence which has experienced a rapid development and has become a powerful tool in a wide range of applications In studies relevant to both evolutionary algorithms and machine learning techniques many attempts have been made to apply variants evolutionary algorithms as types of effective and efficient machine learning techniques In contrast with the view of using evolutionary algorithms as machine learning techniques we have focused on using machine learning techniques to enhance evolutionary algorithms In the framework of enhanced evolutionary algorithms with machine learning techniques the main idea is that the evolutionary algorithm has stored ample data about the search space problem features and population information during the iterative search process thus the machine learning technique is helpful in analyzing these data for enhancing the search performance In this way useful information can be extracted to understand the search behavior and to assist with future searches for the global optimum In many applications evolutionary algorithms incorporating machine learning techniques have been proven to be advantageous in both convergence speed and solution quality The CMA ES algorithm is one of the evolutionary algorithms that produce new population by sampling from a normal distribution Covariance Matrix is one of the normal distribution parameters that is updated in each generation and has an important role in population quality and consequently guidance of the evolution Therefore how updating this covariance matrix is critical On the other hand the DCC GARCH model is one of the machine learning techniques that have several applications in economic models This model predict multivariate time series which each term of the series is sampled from a normal distribution with zero mean and covariance matrix H t Instead of predicting next term directly this model predict the covariance matrix of the next term such that sampling from this covariance matrix may produce next term of time series In this thesis we want to use DCC GARCH model to enhance the CMA ES algorithm The best step in each generation is equal to a term of series and the model will predict the covariance matrix of the best step of next generation by this series Then this matrix is used in updating the covariance matrix of the CMA ES algorithm directly and in directly that the results and experiments is showing reasonable enhancement Key Words Evolutionary Algorithms Machine Learning CMA ES DCC GARCH

عبدالرضا ميرزايي

محمدعلي منتظري

مهران صفاياني، ناصر قديري مدرس