This event will be held virtually.
Speaker: Qi Xiao, ECpE Graduate Student
Advisor: Zhengdao Wang
Title: On Learning Instance-wise Feature Selection, and Its Usage in Supervision-Aware Clustering
Abstract: Feature selection is helpful for understanding data and machine learning models. We consider instance-wise feature selection problem which is more flexible for model interpretation than global feature selection. Existing instance-wise feature selection method does not fit easily on data that follows mixture distribution. Also, the alternative training process of INVASE is expensive. We propose a novel approach using the mixture expert model to learn the feature selector, and we show it works better on heterogeneous data distribution. In the second work, we propose a novel Supervision-Aware Clustering (SAC) problem, and present a simple and effective method based on instance-wise feature selection algorithm. We demonstrate the effectiveness of our method on synthetic and real-world datasets.
Bio: Qi Xiao received her Bachelor degree from University of Science and Technology of China. She is currently a PhD student in the Electrical and Computer Engineering department at Iowa State University. Her research interests include ensemble methods, feature selection and deep learning models.
Webex Link: https://iastate.webex.com/iastate/j.php?MTID=mefb48616fc77aeca05f90274fe7b50fa
Recording: https://iastate.webex.com/iastate/ldr.php?RCID=2381d78cd66c4b388ca51c989beb6011