Issue No.02 - March/April (1998 vol.13)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/5254.671093
The authors present Genetic Algorithm Based Representation Transformation (GABRET), the system they created to transform feature spaces to improve classification techniques. Depending on the problem, the system applies either a feature-selection or ?construction module to search the problem space and improve the recognition rate. Both methods are based on genetic algorithms that use an evaluation function as feedback to guide the search. The authors test this method on an eye-detection face recognition system, demonstrating substantially better classification rates than competing systems.
Genetic Algorithms, machine learning, computer vision, image processing, feature selection, feature construction, feature transformation
Haleh Vafaie, Kenneth De Jong, "Feature Space Transformation Using Genetic Algorithms", IEEE Intelligent Systems, vol.13, no. 2, pp. 57-65, March/April 1998, doi:10.1109/5254.671093