Adding new models to MLJFlux

This section assumes familiarity with the MLJ model API

If one subtypes a new model type as either MLJFlux.MLJFluxProbabilistic or MLJFlux.MLJFluxDeterministic, then instead of defining new methods for MLJModelInterface.fit and MLJModelInterface.update one can make use of fallbacks by implementing the lower level methods shape, build, and fitresult. See the classifier source code for an example.

One still needs to implement a new predict method.