SYSTEM Cited by 1 source
XGBoost¶
XGBoost (eXtreme Gradient Boosting) is a scalable, widely used open-source gradient-boosted decision-tree library authored by Tianqi Chen. It is a common choice for production tabular classification and regression where tree ensembles outperform deep learning — Zalando Payments' deferred-payment risk-scoring being one such instance.
Stub page. Expand as dedicated sources arrive.
Wiki positioning¶
- Named as one of the interchangeable "main model" containers in a SageMaker Inference Pipeline Model alongside PyTorch and TensorFlow (Source: sources/2021-02-15-zalando-a-machine-learning-pipeline-with-real-time-inference).
- Available as a pre-built SageMaker container — Zalando explicitly calls out that the training step can be swapped for any SageMaker-provided algorithm image, with XGBoost being the canonical first choice.
- Zalando had already ingested XGBoost into their production stack in a 2020 blog post on distributed XGBoost on SageMaker; the 2021 real-time inference post references it as an existing in-house option.
Seen in¶
- sources/2021-02-15-zalando-a-machine-learning-pipeline-with-real-time-inference — named as one of three framework options ("XGBoost, PyTorch") for the main-model container stage of the SageMaker Inference Pipeline Model used for deferred-payment fraud / default risk scoring. Zalando Payments previously published on distributed XGBoost on SageMaker in 2020 (referenced in-line but not ingested separately).