Transfer Learning for High-Dimensional Quantile Regression via Convolution Smoothing

• DOI: 10.5705/ss.202022.0396
• Publication Date: 2023-06-26T06:02:16Z
• Authors: Yijiao Zhang; Zhongyi Zhu
• Keywords: Quantile regression; Smoothing; Convolution (computer science); Quantile; Transfer of learning

This paper studies the high-dimensional quantile regression problem under transfer learning framework, where possibly related source datasets are available to make improvements on estimation or prediction based solely target data.In oracle case with known transferable sources, a smoothed two-step algorithm convolution smoothing is proposed and ℓ1/ℓ2 error bounds of corresponding estimator also established.To avoid including non-informative we propose select sources adaptively establish its selection consistency regular conditions.Monte Carlo simulations as well an empirical analysis gene expression data demonstrate effectiveness procedure.