RareXplain: A Computational Framework for Explainable Rare Category Analysis

This project will focus on real-world problems where underrepresented, rare (abnormal) examples play critical roles, such as defective silicon wafers resulting from a new semiconductor manufacturing process and rare but severe complications (e.g., kidney failure) among diabetes patients.

"This problem of explainable rare category analysis was motivated by my collaboration with IBM Research and Mayo Clinic Arizona," said He. "In both semiconductor manufacturing and healthcare, the targets of interest are rare but of great importance. Existing models in this area are mostly black box in nature, making them difficult to be comprehended by domain experts with limited knowledge in AI."

According to He, the new project will bridge the gap between the imminent need to analyze complex rare categories and the inability of state-of-the-art techniques to address this problem in an effective, efficient, and explainable way.

Read more