Genetic and Evolutionary Feature Extraction via X-TOOLSS

In [14], Shelton et al. presented two Genetic and Evolutionary Methods (GEMs) that evolved feature extractors (FE) for facial recognition. One of the methods presented evolved FEs that consisted of non-uniform, overlapping “patches” that did not cover the entire image. The other was similar with the exception that it evolved FEs that consisted of patches that were of uniform size. The two GEMs, referred to as Genetic & Evolutionary Feature Extractors (GEFE), were instances of a Steady-State Genetic Algorithm (SSGA). In [14], the instance of GEFE that evolved FEs consisting of uniform sized patches outperformed its counterpart that evolved FEs consisting of non-uniformed sized patches. This paper compares SSGA instances of GEFE, as reported in [14], with two additional Estimation of Distribution Algorithm (EDA) instances of GEFE. Our results show that the EDA instance of GEFE that evolved FEs consisting of uniform sized patches had the best overall performance.