ganblr.models package

The GANBLR models.

class ganblr.models.GANBLR

Bases: object

The GANBLR Model.

evaluate(x, y, model='lr') → float

Perform a TSTR(Training on Synthetic data, Testing on Real data) evaluation.

Parameters

• x (array_like) – Test dataset.

• y (array_like) – Test dataset.

• model (str or object) – The model used for evaluate. Should be one of [‘lr’, ‘mlp’, ‘rf’], or a model class that have sklearn-style fit and predict method.

• Return –

• -------- –

• accuracy_score (float.) –

fit(x, y, k=0, batch_size=32, epochs=10, warmup_epochs=1, verbose=1)

Fit the model to the given data.

Parameters

• x (array_like of shape (n_samples, n_features)) – Dataset to fit the model. The data should be discrete.

• y (array_like of shape (n_samples,)) – Label of the dataset.

• k (int, default=0) – Parameter k of ganblr model. Must be greater than 0. No more than 2 is Suggested.

• batch_size (int, default=32) – Size of the batch to feed the model at each step.

• epochs (int, default=0) – Number of epochs to use during training.

• warmup_epochs (int, default=1) – Number of epochs to use in warmup phase. Defaults to 1.

• verbose (int, default=1) – Whether to output the log. Use 1 for log output and 0 for complete silence.

Returns

self – Fitted model.

Return type

object

sample(size=None, verbose=1) → ndarray

Generate synthetic data.

Parameters

• size (int or None) – Size of the data to be generated. set to None to make the size equal to the size of the training set.

• verbose (int, default=1) – Whether to output the log. Use 1 for log output and 0 for complete silence.

• Return –

• ----------------- –

• synthetic_samples (np.ndarray) – Generated synthetic data.

class ganblr.models.GANBLRPP(numerical_columns, random_state=None)

Bases: object

The GANBLR++ model.

Parameters

• numerical_columns (list of int) – Indicating the indexes of numerical columns. For example, if the 3, 5, 10th feature of a data is numerical feature, then this param should be [3, 5, 10].

• random_state (int, RandomState instance or None) – Controls the random seed given to the method chosen to initialize the parameters of BayesianGaussianMixture used by GANBLRPP.

evaluate(x, y, model='lr')

Perform a TSTR(Training on Synthetic data, Testing on Real data) evaluation.

Parameters

• x (array_like) – Test dataset.

• y (array_like) – Test dataset.

• model (str or object) – The model used for evaluate. Should be one of [‘lr’, ‘mlp’, ‘rf’], or a model class that have sklearn-style fit and predict method.

• Return –

• -------- –

• accuracy_score (float.) –

fit(x, y, k=0, batch_size=32, epochs=10, warmup_epochs=1, verbose=1)

Fit the model to the given data.

Parameters

• x (array_like of shape (n_samples, n_features)) – Dataset to fit the model. The data should be discrete.

• y (array_like of shape (n_samples,)) – Label of the dataset.

• k (int, default=0) – Parameter k of ganblr model. Must be greater than 0. No more than 2 is Suggested.

• batch_size (int, default=32) – Size of the batch to feed the model at each step.

• epochs (int, default=0) – Number of epochs to use during training.

• warmup_epochs (int, default=1) – Number of epochs to use in warmup phase. Defaults to 1.

• verbose (int, default=1) – Whether to output the log. Use 1 for log output and 0 for complete silence.

Returns

self – Fitted model.

Return type

object

sample(size=None, verbose=1)

Generate synthetic data.

Parameters

• size (int or None) – Size of the data to be generated. set to None to make the size equal to the size of the training set.

• verbose (int, default=1) – Whether to output the log. Use 1 for log output and 0 for complete silence.

• Return –

• ----------------- –

• synthetic_samples (np.ndarray) – Generated synthetic data.