Table of Contents

Mojo struct

GBDT

@memory_only
struct GBDT

Gradient Boosting with support for both classification and regression.

Aliases

  • __del__is_trivial = False

Fields

  • criterion (String): The method to measure the quality of a split: For binary classification -> 'log'; For multi-class classification -> 'softmax'; For regression -> 'mse'.
  • loss_g (fn(Matrix, Matrix) raises -> Matrix)
  • loss_h (fn(Matrix) raises -> Matrix)
  • n_trees (Int): The number of boosting stages to perform.
  • min_samples_split (Int): The minimum number of samples required to split an internal node.
  • max_depth (Int): The maximum depth of the tree.
  • learning_rate (Float32): Learning rate.
  • reg_lambda (Float32): The L2 regularization parameter.
  • reg_alpha (Float32): The L1 regularization parameter.
  • gamma (Float32): Minimum loss reduction required to make a further partition on a leaf node of the tree.
  • n_bins (Int): Generates histogram boundaries as possible threshold values when n_bins >= 2 instead of all possible values.
  • trees (UnsafePointer[BDecisionTree, MutAnyOrigin])
  • score_start (Float32)
  • num_class (Int)

Implemented traits

AnyType, CV, ImplicitlyDestructible

Methods

__init__

fn __init__(out self, criterion: String = "log", n_trees: Int = 10, min_samples_split: Int = 10, max_depth: Int = 3, learning_rate: Float32 = 0.10000000000000001, reg_lambda: Float32 = 1, reg_alpha: Float32 = 0, gamma: Float32 = 0, n_bins: Int = 0)

Args:

  • criterion (String)
  • n_trees (Int)
  • min_samples_split (Int)
  • max_depth (Int)
  • learning_rate (Float32)
  • reg_lambda (Float32)
  • reg_alpha (Float32)
  • gamma (Float32)
  • n_bins (Int)
  • self (Self)

Returns:

Self

fn __init__(out self, params: Dict[String, String])

Args:

  • params (Dict)
  • self (Self)

Returns:

Self

Raises:

__del__

fn __del__(deinit self)

Args:

  • self (Self)

fit

fn fit(mut self, X: Matrix, y: Matrix)

Fit the gradient boosting model.

Args:

  • self (Self)
  • X (Matrix)
  • y (Matrix)

Raises:

predict

fn predict(self, X: Matrix) -> Matrix

Predict class or regression value for X.

Args:

  • self (Self)
  • X (Matrix)

Returns:

Matrix: The predicted values.

Raises:

Edit this page