Differential Evolution with Generalized Mutation Operator for Parameters Optimization in Gene Selection for Cancer Classification
This work addresses the need for more effective evolutionary algorithms for continuous optimization, with applications in bioinformatics for cancer classification, but it appears incremental as it builds on existing DE methods.
The authors tackled the problem of improving Differential Evolution (DE) for global optimization by proposing a new notation and four novel mutations, leading to the Generalized Mutation Differential Evolution (GMDE) algorithm. Results on CEC2005 benchmarks showed GMDE is surprisingly competitive and significantly improved performance, and it was applied to optimize parameters for gene selection in cancer classification.
Differential Evolution (DE) proved to be one of the most successful evolutionary algorithms for global optimization purposes in continuous problems. The core operator in DE is mutation which can provide the algorithm with both exploration and exploitation. In this article, a new notation for DE is proposed which has a formula that can be utilized for generating and extracting novel mutations and by applying this new notation, four novel mutations are proposed. More importantly, by combining these novel trial vector generation strategies and four other well-known ones, we proposed Generalized Mutation Differential Evolution (GMDE) that takes advantage of two mutation pools that have both explorative and exploitative strategies inside them. Results and experimental analysis are performed on CEC2005 benchmarks and the results stated that GMDE is surprisingly competitive and significantly improved the performance of this algorithm. Finally, GMDE is also applied to parameters optimization, modification and improvement of a feature selection method for cancer classification purposes over gene expression microarray profiles.