Register a model backend — register_model

Add or replace a model backend in the registry. Once registered, the backend name appears in the model-type dropdown and the custom-model-lines pipeline dispatches through it automatically.

Usage

register_model_backend(name, backend)

Arguments

name: Character string. The name that will appear in the model-type dropdown (e.g. "drm", "gam").
backend: A named list with elements fit, predict, and validate_classes as described above.

Value

Invisibly returns NULL. Called for its side effect.

Details

A backend is a named list with three required elements:

fit: A function with signature function(formula, data, ...) that returns a fitted model object. Extra UI fields from the multiDynamicInput() row are forwarded as ....
predict: A function with signature function(model, newdata) that returns a numeric vector of predicted y-values, one per row of newdata.
validate_classes: Character vector of class names. After fitting, the pipeline checks inherits(model, validate_classes) and rejects the model if it fails.

Author

Jacob Martin

Examples

# Register a custom backend for dose-response curves (requires drc)
if (requireNamespace("drc", quietly = TRUE)) {
}
#> NULL