Global Feature Subset Selection on High-Dimensional Datasets Using Re-ranking-based EDAs

The relatively recent appearance of high-dimensional databases has made traditional search algorithms too expensive in terms of time and memory resources. Thus, several modifications or enhancements to local search algorithms can be found in the literature to deal with this problem. However, nondeterministic global search, which is expected to perform better than local, still lacks appropriate adaptations or new developments for high-dimensional databases. We present a new non-deterministic iterative method which performs a global search and can easily handle datasets with high cardinality and, furthermore, it outperforms a wide variety of local search algorithms.