generic split stat.

[Julia Br]

Добрый день!
Помогите, пожалуйста, разобраться с переводом алгоритма построения дерева решений. Что это может быть за разбиение?

It is often more desirable to form a tree whose root node is split on a X descriptor (a generic split). This is because a generic split at the root would better utilize training items. For example, a training item set may consist of 1000 training items with known property Y1 and 20 training items with known property Y2. A generic split at the root uses the 1020 training items to split. If the root node is split by K, then a subsequent generic split by Y2 would only use the 20 training items with known property Y2.

Подробнее про X descriptor:
Application describes systems and methods of forming a decision tree that simultaneously classifies multiple properties of items.
Training items having multiple X descriptors and multiple known properties are converted into multiple converted items of the same multiple X descriptors, an additional property type descriptor (called the K descriptor), and one property. The maximally generic tree is created by first splitting the root node on an X descriptor and splitting its immediate descendent nodes on the K descriptor. The final decision tree produced with this method has a top generic layer of nodes that are generic to all properties, a middle K layer of nodes that split on descriptor K, and a bottom specific layer of nodes that are specific to some properties.

https://worldwide.espacenet.com/publicationDetails/description?CC=CA&NR=2364410A1&KC=A1&FT=D&ND=&date=20020703&DB=&locale=

Буду очень признательна за помощь!

[I. Havkin]

Есть generalized splitting - обобщённое разбиение (расщепление). Может быть, generic split - это то же самое?
См., напр., http://www.math.uni-bielefeld.de/LAG/man/050.pdf

[Julia Br]

Да, похоже. Большое спасибо!!!