On the Complexity of One-class SVM for Multiple Instance Learning

Zhen Hu, Zhuyin Xue

In traditional multiple instance learning (MIL), both positive and negative bags are required to learn a prediction function. However, a high human cost is needed to know the label…