> ML_LITERATURE // KE-2017-LIGHTGBM-HIGHLY-EFFICIENT-GBDT_v1.0
LightGBM: A Highly Efficient Gradient Boosting Decision Tree
Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, Tie-Yan Liu · Advances in Neural Information Processing Systems (NeurIPS) (2017)
systems2017industry-standardthirdPartyReproduced
Principal Contribution
Introduced Gradient-based One-Side Sampling (GOSS) and Exclusive Feature Bundling (EFB) alongside leaf-wise tree growth.
Operational Relevance
Industry-standard choice for low-latency training and high-throughput scoring on large-scale tabular production datasets.
Assumptions
- Data instances with small gradients contribute less to training and can be subsampled without harming accuracy
Limitations
- Leaf-wise tree growth can overfit rapidly on small datasets (<10k rows) unless max_depth is explicitly restricted
Connected Algorithms, Architectures & Tools
Related Algorithms:
Related Architectures:
Implementing Libraries: