tabensemb.trainer.Trainer.cross_validation#

method

Trainer.cross_validation(programs: List[str], n_random: int, verbose: bool, test_data_only: bool, split_type: str = 'cv', load_from_previous: bool = False, **kwargs) Dict[str, Dict[str, Dict[str, Tuple[ndarray, ndarray]]]][source]#

Repeat load_data(), train model bases, and evaluate all models for multiple times.

Parameters:
programs

A selected subset of model bases.

n_random

The number of repeats.

verbose

Verbosity.

test_data_only

Whether to evaluate models only on testing datasets.

split_type

The type of data splitting. “random” and “cv” are supported. Ignored when load_from_previous is True.

load_from_previous

Load the state of a previous run (mostly because of an unexpected interruption).

**kwargs

Arguments for tabensemb.model.AbstractModel.train()

Returns:
dict

A dict in the following format: {keys: programs, values: {keys: model names, values: {keys: [“Training”, “Testing”, “Validation”], values: (Predicted values, true values)}}

Notes

The results of a continuous run and a continued run (load_from_previous=True) are consistent.