nsl.estimator.add_adversarial_regularization

Adds adversarial regularization to a tf.estimator.Estimator.

nsl.estimator.add_adversarial_regularization(
    estimator, optimizer_fn=None, adv_config=None
)

The returned estimator will include the adversarial loss as a regularization term in its training objective, and will be trained using the optimizer provided by optimizer_fn. optimizer_fn (along with the hyperparameters) should be set to the same one used in the base estimator.

If optimizer_fn is not set, a default optimizer tf.train.AdagradOptimizer with learning_rate=0.05 will be used.

estimator A tf.estimator.Estimator object, the base model.
optimizer_fn A function that accepts no arguments and returns an instance of tf.train.Optimizer. This optimizer (instead of the one used in estimator) will be used to train the model. If not specified, default to tf.train.AdagradOptimizer with learning_rate=0.05.
adv_config An instance of nsl.configs.AdvRegConfig that specifies various hyperparameters for adversarial regularization.

A modified tf.estimator.Estimator object with adversarial regularization incorporated into its loss.