Capability table
Optional engines remain optional. Missing packages are reported explicitly rather than changing the scientific task or silently selecting another model.
capabilities <- gp3ml_engine_capabilities(
check_keras_backend = FALSE
)
capabilities
#> engine package classification regression probability package_available
#> glm <NA> TRUE FALSE TRUE TRUE
#> lm <NA> FALSE TRUE FALSE TRUE
#> ranger ranger TRUE TRUE TRUE TRUE
#> xgboost xgboost TRUE TRUE TRUE TRUE
#> nnet nnet TRUE TRUE TRUE TRUE
#> keras3 keras3 TRUE TRUE TRUE TRUE
#> custom <NA> TRUE TRUE NA TRUE
#> backend backend_ready status
#> <NA> NA available
#> <NA> NA available
#> <NA> NA available
#> <NA> NA available
#> <NA> NA available
#> <NA> NA backend_unverified
#> <NA> NA available
#> notes
#> Base-R binomial GLM.
#> Base-R linear model.
#> Optional package; governed wrapper.
#> Optional package; governed wrapper.
#> Recommended R package; governed wrapper.
#> Optional package plus configured backend; deep learning remains explicit.
#> Externally supplied engine requires safety declarations.
plot(capabilities)
glm and lm are always available.
ranger, xgboost, nnet, and
keras3 are exercised by a dedicated GitHub Actions matrix.
Keras backend readiness is queried only when explicitly requested.
Engine availability is not a model-selection rule. Candidate selection remains explicit, metric-declared, direction-declared, and reviewable.