Originally published at: Security for Data Privacy in Federated Learning with CUDA-Accelerated Homomorphic Encryption in XGBoost | NVIDIA Technical Blog
XGBoost is a machine learning algorithm widely used for tabular data modeling. To expand the XGBoost model from single-site learning to multisite collaborative training, NVIDIA has developed Federated XGBoost, an XGBoost plugin for federation learning. It covers vertical collaboration settings to jointly train XGBoost models across decentralized data sources, as well as horizontal histogram-based and…