Granular ball computing classifiers for efficient, scalable and robust learning

Shuyin Xia, Yunsheng Liu, Xin Ding, Guoyin Wang, Hong Yu, Yuoguo Luo

Abstract: Highlights•Granular ball computing classifier (GBC) is proposed as a framework by introducing granular balls into existing classifiers to make them efficient, scalable and robust.•Granular ball support vector machine (GBSVM) and granular ball k-nearest neighbors’ algorithms (GBkNN) are derived. The GBC almost have a time complexity of nearly O(n).•The algorithm of the proposed granular ball kNN (GBkNN) is designed.•The existing classifiers can be seen as the finest versions of GBC in which each granular ball contains one point, and the GBC can lead to better generalizability on some cases.