Mingrui Liu, Department of Computer Science, The University of Iowa
601 Pao Yue-Kong Library
Online learning receives tremendous attention since it can handle streaming data. In many applications (e.g., medical diagnostics, spam email detection, malicious URL detection), we are facing with imbalanced data where the number of positive samples is much larger than the number of negative samples. Classical optimization algorithms designed for minimizing the misclassification rate are not suitable for handling large-scale imbalanced data. In this talk, I will present a stochastic optimization algorithm for optimizing AUC (Area under the ROC Curve). Our proposed algorithm improves over the state-of-the-art algorithms in terms of computational complexity, and also shows better performance in real applications.