privpack.model_selection.experiment module

Some introduction on how to define experiments.

API Documentation

This module defines how experiments are conducted. What is expected from your network and how many train/validation splits should be trained and tested.

class privpack.model_selection.experiment.Expectations

Bases: object

class Expectation(metric: privpack.utils.metrics.Metric, value: float, relation: Callable[[float, float], bool] = <function Expectations.Expectation.<lambda>>)

Bases: object

add_expectation(metric: privpack.utils.metrics.Metric, value: float, relation: Callable[[float, float], bool] = <function Expectations.<lambda>>)
class privpack.model_selection.experiment.Experiment(network: privpack.core.architectures.GenerativeAdversarialNetwork, expectations: privpack.model_selection.experiment.Expectations)

Bases: object

run(data, n_splits, epochs, batch_size, verbose=False, **kwargs)