Zhou Jing

In order to further promote the School's international cooperation and exchange, the School of Mathematics, Physicsand Statistics invited Associate Professor Su Zhihua from the University of Florida to attend the 2024 International Conference on Optimization and Statistics with its Applications online on June 22nd. Assoc. Prof. Su gave an academic report entitled "Envelope-basedpartial least squares".

 Assoc. Prof. Su introduced that compared with ordinary least squares (OLS), partialleast squares (PLS) usually has better prediction performance, and the PLS algorithm can be directly applied to the case of n<p. She went on to point out that although PLS is very popular, the theoretical properties of its estimators are still largely unknown. As a result, it is difficult to determine when PLS is superior to OLS, what the limits of PLS are, and how PLS can be improved. Assoc. Prof. Su mainly explained her research results on three issues in PLS, including variable selection, classification predictors and scale invariance, respectively.

At the end of the report, Associate Professor Zhihua Su interacted with experts and students present and answered their questions about partial least square method and its application in statistics in detail.