@article{1489,
  author = {Azra Aryania, Ahmad Akbari, Mahdi Mohammadi, Bijan Raahemi, Elnaz Bigdeli},
  title = {An Overlap-aware Positive Selection Algorithm using Variable-size Detectors},
  journal = {Journal of Intelligent Computing},
  year = {2014},
  volume = {5},
  number = {2},
  doi = {},
  url = {http://www.dline.info/jic/fulltext/v5n2/3.pdf},
  abstract = {Classification of the test samples using positive selection is computationally expensive, as it requires comparisons with large number of detectors. In this paper, we propose an enhanced positive selection algorithm with variable-size detectors to reduce the number of detectors required to cover the training sample space while resolving the issues of ambiguity of the test samples (test samples covered by more than one detector from different classes) and exclusion (test samples not covered by any detectors). We apply clustering in the preprocessing phase to reduce the number of detectors. We then perform overlap checking to adjust the radiuses of the variable-size detectors aiming at reducing overlap while covering the sample space. Furthermore, a weighted voting scheme is employed to resolve the ambiguity, and a distance-based method is devised to resolve the issue of exclusion. We evaluate the performance of our proposed algorithm on five benchmark datasets. The experimental results confirm the superiority of the proposed method in terms of number of detectors and accuracy rate.},
}