평균 이동 알고리즘을 이용한 지지 벡터 영역 표현 학습시간 단축 방법

Quadratic programming solver of Support Vector Data Description (SVDD) has a runtime complexity of O(N³). So SVDD has a limitation of dealing with a large data set. To handle this scale problem, we propose SVDD using Mean Shift clustering method, which is finding the modes of data distribution first and then clustering the data based on the modes. This algorithm’s computational time reduction performance is outstanding, and we can get the same results which are containing the characteristic of data distribution regardless of how many times we carried out the experiments. Also Mean Shift window bandwidth has some margin in deciding.