XGBoost Introduces Vector-Leaf Model
New tree structure stores vectors in leaves to support multiple predictions per row.
Tianqi Chen posted a link to the XGBoost site announcing the Vector-Leaf Model created by Jiamin Yuan and Rory Mitchell. The post describes standard decision trees that hold one number per leaf for single-value regression. It notes that some tasks need several related outputs from the same row. The Vector-Leaf Model changes leaf contents to hold vectors instead. The announcement appears on the official XGBoost project site and is presented as the latest update to the library.
Combined views
74.1K
3 posts, first seen 23h ago
Reactions
3 postsTC 
Tianqi Chen@tqchenmlCheckout the latest XGBoost update "Introducing the XGBoost Vector-Leaf Model" from jiamin yuan and Rory Mitchell
- likes: 296
- replies: 3
- bookmarks: 237
- reposts: 42
J🇫🇷 JFPuget 🇫🇷🇺🇦🇨🇦🇬🇱@JFPugetSignificant: one can now use XGBoost for multi class or multi label problems without having to train a model for each class or label.
- likes: 253
- replies: 5
- bookmarks: 150
- reposts: 17
J🇫🇷 JFPuget 🇫🇷🇺🇦🇨🇦🇬🇱@JFPugetSignificant: one can now use XGBoost for multi class or multi label problems without having to train a model for each class or label.
- likes: 253
- replies: 5
- bookmarks: 150
- reposts: 17