treeboost 0.1.0

High-performance Gradient Boosted Decision Tree engine for large-scale tabular data
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

// Structure
pub const DEFAULT_MAX_DEPTH: usize = 6;
pub const DEFAULT_MAX_LEAVES: usize = 31;
pub const SHALLOW_MAX_DEPTH: usize = 4;
pub const DEEP_MAX_DEPTH: usize = 10;

// Regularization
pub const DEFAULT_TREE_LAMBDA: f32 = 1.0;
pub const REGULARIZED_TREE_LAMBDA: f32 = 2.0;
pub const EXPRESSIVE_TREE_LAMBDA: f32 = 0.0;
pub const DEFAULT_ENTROPY_WEIGHT: f32 = 0.0;
pub const REGULARIZED_ENTROPY_WEIGHT: f32 = 0.3;
pub const DEFAULT_MIN_GAIN: f32 = 0.0;

// Samples
pub const DEFAULT_MIN_SAMPLES_LEAF: usize = 1;
pub const DEFAULT_MIN_HESSIAN_LEAF: f32 = 1.0;
pub const DEFAULT_COLSAMPLE: f32 = 1.0;

// Step size
pub const DEFAULT_LEARNING_RATE: f32 = 0.1;