Reading queue: XGBoost vector-leaf models
Introducing the XGBoost Vector-Leaf Model is an August 25 technical article from the XGBoost project, saved to my reading queue on September 4.
Reading angle
How does sharing tree structure across outputs change the tradeoff between model quality, training cost, and model size? I want to compare that design with independent output models in the tabular foundation model lab.
This is a queued reading and experiment direction; I have not recorded benchmark results for it here.