Cortex AI logo placeholderCortexLabAI · learning · engineering
Model Atlas · Boosted trees

Gradient Boosting

Sequential trees correct residual errors and often dominate structured-data benchmarks.

core mathematical viewFₘ(x)=Fₘ₋₁(x)+ηhₘ(x)
Mental model

Understand it before memorizing it.

Each new weak learner focuses on what the current ensemble still gets wrong.

Best fit

Where this model earns its place

High-quality tabular prediction

Start with a simpler baseline then compare this model using the same evaluation split and operational constraints.

Ranking

Start with a simpler baseline then compare this model using the same evaluation split and operational constraints.

Complex interactions

Start with a simpler baseline then compare this model using the same evaluation split and operational constraints.

Strengths and limits

Trade-offs matter more than popularity.

Strengths

✓ Strong accuracy

✓ Flexible losses

✓ Handles nonlinearities

Limitations

△ Sensitive tuning

△ Can overfit noisy data

Evaluation

Metrics to watch

PR-AUCInterpret with the product objective and error cost.
Log lossInterpret with the product objective and error cost.
RMSEInterpret with the product objective and error cost.
NDCGInterpret with the product objective and error cost.
Production checklist

Before it reaches real users

  1. 01

    Tune learning rate/depth

    Document the assumption and instrument the condition so regressions can be detected.

  2. 02

    Watch drift

    Document the assumption and instrument the condition so regressions can be detected.

  3. 03

    Use early stopping

    Document the assumption and instrument the condition so regressions can be detected.