- Base Leave-k-out - trained_models dictionary field -> {idx/tuples: model} - `add_model(...)` method in the base -> adds models to trained_models - `update(...)`_can only calculate the metric w.r.t. the indices of the trained models_ - cache the re-trained models for re-use - think about leave-k-out "benchmarks" to be uploaded to the cloud
add_model(...)method in the base -> adds models to trained_modelsupdate(...)can only calculate the metric w.r.t. the indices of the trained models