params |
Object
|
|
Dictionary with parameters for
the model.
Properties
| Name |
Type |
Attributes |
Description |
loss |
String
|
<optional>
|
Type of loss to use. Currently
available is 'hinge'. |
|
param.learning_rate |
Number
|
<optional>
|
How much does a single
weak learner affect the final ensemble. Smaller values might
require larger number of estimators for satisfactory behavior,
but usually lead to better generalization. |
param.n_estimators |
Number
|
<optional>
|
How many weak learners
to use in final ensemble. |
param.min_samples_split |
Number
|
<optional>
|
Hyperparameter of
the underlying weak learner. Minimum number of samples in the
decision tree to form a split. Can either be a value in range
[0.0, 1.0], which indicates the fraction of overall dataset.
Alternatively, it could be a value greater equal 2, in which
case specific number of samples is specified. |
param.max_depth |
Number
|
<optional>
|
Maximum depth of the decision
tree weak learner. |
param.min_impurity_decrease |
Number
|
<optional>
|
Minimal decrease
of impurity value, for which to form a new split. Used in the
training heuristic of the tree weak learner. Recommended to leave
as 0.0. |