A Novel Incremental Covariance-Guided One-Class Support Vector Machine

Covariance-guided One-Class Support Vector Machine (COSVM) is a very competitive kernel classifier, as it emphasizes the low variance projectional directions of the training data, which results in high accuracy. However, COSVM training involves solving a constrained convex optimization problem, which requires large memory and enormous amount of training time, especially for large scale datasets. Moreover, it has difficulties in classifying sequentially obtained data. For these reasons, this paper introduces an incremental COSVM method by controlling the possible changes of support vectors after the addition of new data points. The control procedure is based on the relationship between the Karush-Kuhn-Tuker conditions of COSVM and the distribution of the training set. Comparative experiments have been carried out to show the effectiveness of our proposed method, both in terms of execution time and classification accuracy. Incremental COSVM results in better classification performance when compared to canonical COSVM and contemporary incremental one-class classifiers.