Module multi_layer 

Multi-Layer Stacking Ensemble Implementation

This module provides advanced multi-layer stacking ensemble methods that combine multiple layers of base estimators with sophisticated meta-learning strategies and feature engineering. The implementation supports:

• Deep stacking with multiple layers
• Advanced meta-feature engineering strategies
• Ensemble pruning and diversity analysis
• Confidence-based weighting
• Multiple meta-learning strategies
• SIMD-accelerated operations for performance

Features

Multi-Layer Architecture

• Configurable number of stacking layers
• Layer-wise meta-feature generation
• Hierarchical feature transformation

Advanced Meta-Feature Engineering

• Statistical features (mean, std, skewness, etc.)
• Interaction features (pairwise products)
• Confidence-based features (entropy, agreement)
• Diversity-based features (coefficient of variation)
• Comprehensive features (combination of all strategies)
• Temporal features for time-series data
• Spectral features using FFT analysis
• Information-theoretic features (mutual information, entropy)
• Neural embedding features
• Kernel-based features (RBF, polynomial, cosine)
• Basis expansion features (Legendre polynomials)
• Meta-learning features (complexity, stability, agreement)

Ensemble Optimization

• Diversity-based ensemble pruning
• Layer-wise feature importance analysis
• Confidence weighting
• Multiple regularization strategies

Example

use sklears_ensemble::stacking::multi_layer::MultiLayerStackingClassifier;
use sklears_ensemble::stacking::config::MultiLayerStackingConfig;
use sklears_core::traits::Fit;
use scirs2_core::ndarray::array;

// Create a deep stacking classifier with 3 layers
let config = MultiLayerStackingConfig::deep_stacking(3, 5);
let classifier = MultiLayerStackingClassifier::new(config);

// Training data
let x = array![[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]];
let y = array![0, 1, 0];

// Fit the model
let fitted = classifier.fit(&x, &y).unwrap();

// Make predictions
let predictions = fitted.predict(&x).unwrap();
let probabilities = fitted.predict_proba(&x).unwrap();

Structs

MultiLayerStackingClassifier

Multi-Layer Stacking Classifier

StackingLayer

Represents a single layer in the multi-layer stacking ensemble