MachineLearning/lightgbm_like library

🌩 LightGBM-like Regressor (simplified)

Simplified LightGBM-style regressor that demonstrates key ideas: feature subsampling, leaf-wise growth via greedy stump selection, and fast split enumeration using unique feature values. This is educational, not a production implementation.

Contract:

  • Input: X (n x m), y numeric targets.
  • Output: LightGBMLikeRegressor with fit and predict.

Time Complexity: O(n * m * n_estimators) Space Complexity: O(n_estimators)