An XGboost Decision Tree Note that algorithms are independent of Nodes implementation as Node is abstract
Dense Feature vector
Dense Feature vector
features data
considers 0.0 as N/A
A generic Feature Vector
the model Parameters
the model Parameters
base score
number of features
number of classes
number of trees
numbet of groups
objective name
A generic (binary) node of a decision Tree
The node implementation (actually a simple case class)
The node implementation (actually a simple case class)
left child
right child
rw node id (just used for information)
split index
is left child the default
node value
sum hess
mean value
The XGboost Objective functions
Element of a unique Path in SHAP algorithm https://arxiv.org/pdf/1802.03888.pdf
Predictor fox XGboost
Predictor fox XGboost
model parameters
decision tree
objective function
Builder to ease the creation of Predictor from Java (avoid using Scala Module name)
RawNode is just use to read the trees from binary model file
RawNode is just use to read the trees from binary model file
node id
left node id
right node id
is this a leaf node
leaf value or split value
is left child the default
split index
sum Hess
Sparse Feature vector
Sparse Feature vector
a map of position/value
DenseFVector companion object is used to handle Java collections
SparseFVector companion object is used to handle Java collections
An XGboost Decision Tree Note that algorithms are independent of Nodes implementation as Node is abstract
root node