autoxgbparset: autoxgboost default parameter set.

In ja-thomas/autoxgboostMC: Automatic Preprocessing, Fitting and Tuning for xgboost using multiple criteria.

Description

This is the default parameter set for xgboost that is used for tuning if no costum parameter set is defined by the user. For a documentation of the parameter see https://github.com/dmlc/xgboost/blob/master/doc/parameter.md. By default this set is used:

eta

Learning rate: Between 0.01 and 0.2.

gamma

Minimum loss required for a split: Between 2^-7 = 0.0078125 and 2^6 = 64.

max_depth

Maximum tree depth: Between 3 and 20.

colsample_bytree

Subsample ratio of columns per tree: Between 0.5 and 1.

colsample_bylevel

Subsample ratio of clumns per tree level: Between 0.5 and 1.

lambda

L2 Regularization: Between 2^-10 = 0.00097... and 2^10 = 1024.

alpha

L1 Regularization: Between 2^-10 = 0.00097... and 2^10 = 1024.

subsample

Subsample fraction: Between 0.5 and 1.

ja-thomas/autoxgboostMC documentation built on May 17, 2019, 4:22 a.m.