liveBook · Manning

Generalized additive model

As explained in Ensemble Methods for Machine Learning

A generalized additive model is a model whose prediction is formed by adding separate learned functions of input features, optionally including specified interactions.

A generalized additive model (GAM) forms a nonlinear model from separate learned functions, with each component depending on one feature. This preserves an additive, glass-box structure: individual feature effects can be inspected directly instead of reconstructed with a post-hoc explanation method. section 9.5.1 section 9.5

A GAM is useful when the model should remain understandable while capturing nonlinear effects:

  • Treat each feature as its own learned component rather than requiring one opaque model to represent every relationship. section 9.5.1
  • Extend a generalized linear model by replacing linear feature terms with nonlinear functions. section 9.5 section 9.5.1
  • Inspect learned functions to visualize how each feature relates to the target. section 9.5.1
  • Add selected pairwise functions when important interactions are needed; this produces the extended structure used by explainable boosting machines (EBMs). section 9.5 section 9.6

Diabetes diagnosis

An EBM for diabetes diagnosis can contain separate nonlinear components for age, blood glucose level, and body mass index, plus one selected pairwise component. It avoids including every possible feature pair, limiting model bloat while retaining a significant interaction. section 9.5.1

How Ensemble Methods for Machine Learning covers it

How deep it goes
Defines single-feature nonlinear components and selected interactions, then sketches cyclic boosting without deriving the full training procedure.
Examples
Uses diabetes diagnosis with age, blood glucose level, body mass index, and one selected feature pair.
What it assumes
Builds from generalized linear models and assumes familiarity with ensemble learning and gradient boosting.
What you can do after
Evaluate nonlinear feature effects and inspect selected interactions directly through the learned component functions.
How it connects
Connects generalized additive models with generalized linear models, gradient boosting, feature interactions, and model interpretability.
Cover of Ensemble Methods for Machine Learning
liveBookmanning.com
Read Ensemble Methods for Machine Learning on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Ensemble Methods for Machine Learning. The section links point to the parts of the book it draws on.