Exclusive feature bundling is a feature-dimensionality-reduction technique that groups sparse, mutually exclusive features while preserving their nonzero information.
Exclusive feature bundling (EFB) reduces the column dimension of sparse, heavily one-hot-encoded or binary data by grouping features while retaining their nonzero values. LightGBM uses it to improve computational efficiency without compromising data integrity. section 5.5.2
Combining two sparse features
For features A and B, add B to A only at nonzero positions, using the maximum value in A to separate their encoded ranges. Values from each feature remain distinguishable, preserving the original feature order. section 5.5.2
Ensemble Methods for Machine LearningFeature bundling is a tree-learning optimization that groups compatible features to reduce the effective number of features processed.
LightGBMused by · LightGBM is an open-source gradient-boosting framework for efficiently training tree-based machine-learning models.
one-hot encodingcontrasts with · One-hot encoding represents each category as a binary vector or set of indicators with exactly one active component identifying that category.
dimensionality reductionimplements · Dimensionality reduction is the transformation of data into a representation with fewer dimensions while retaining selected information, structure, or relationships.
greedy algorithmimplemented by · A greedy algorithm builds a solution incrementally by choosing the locally best available option at each step, without generally accounting for all future consequences or revising prior choices.
This explainer was generated by Manning's knowledge graph from the text of Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI. The section links point to the parts of the book it draws on.
