About
Welcome to my homepage! I am Peize Wang at the Institute of Statistics and Big Data, Renmin University of China.
My research interests focus on optimal transport and sufficient dimension reduction, with a particular interest in kernel methods for nonlinear representation learning.
At Huawei, I worked on statistical tools for data-quality diagnosis and model-free predictability assessment in key energy-sector problems, including arc-fault and liquid-leakage scenarios.
Promoted the landing of Fano-inversion based tools for estimating error bounds and accuracy ceilings, helping teams distinguish data information limits from model-selection bottlenecks before expensive model tuning.
Built model-free feature-selection workflows based on multivariate mutual information, POTD, and Conditional Wasserstein Dependency Measure (CWDM), targeting high-dimensional, noisy, and redundant industrial data under memory constraints.
Supported early validation on key energy challenges such as arc-fault and liquid-leakage fault diagnosis, turning statistical upper-bound estimation and feature-ranking methods into reusable evaluation pipelines.
I participated in the Zijin Summit young scholars' paper sharing session at Huawei Huang Danian Chasiwu.
Our papers KPOTD: Kernel Principal Optimal Transport Directions for Nonlinear Sufficient Dimension Reduction and SSP-Ensemble: A Sufficient Subspace Projection Ensemble for Multiclass Classification were accepted by the inaugural STAI-X conference.
STAI-X, 2026
Kernelizes optimal-transport displacement covariance in an RKHS to recover nonlinear sufficient dimension reduction structures.
STAI-X, 2026
Builds a supervised POTD-based ensemble by viewing multiclass classification through class-pair sufficient subspace projections.
Please feel free to contact me via email.
wpz2024@ruc.edu.cn