tlda.SecondOrderCumulant

class tlda.SecondOrderCumulant(n_eigenvec, alpha_0, batch_size)[source]

Class to compute the third order cumulant

Methods

fit(X)	Method to fit the entire data to get the projection weights (singular vectors) and whitening weights (scaled explained variance) of a centered input dataset X.
partial_fit(X_batch)	Fit a batch of data and update the projection weights (singular vectors) and whitening weights (scaled explained variance) accordingly using a centered batch of the input dataset X.
reverse_transform(X)	Unwhiten some whitened tensor X using the fitted PCA model.
transform(X)	Whiten some centered tensor X using the fitted PCA model.

fit(X)[source]

Method to fit the entire data to get the projection weights (singular vectors) and whitening weights (scaled explained variance) of a centered input dataset X.

Parameters:

Xtensor of shape (n_samples, vocabulary_size)

Tensor containing all input documents

partial_fit(X_batch)[source]

Fit a batch of data and update the projection weights (singular vectors) and whitening weights (scaled explained variance) accordingly using a centered batch of the input dataset X.

Parameters:

X_batchtensor of shape (batch_size, vocabulary_size)

Tensor containing a batch of input documents

transform(X)[source]

Whiten some centered tensor X using the fitted PCA model.

Parameters:

Xtensor of shape (batch_size, vocabulary_size)

Batch of centered samples

Returns:

whitened_Xtensor of shape (batch_size, self.n_eigenvec)

Whitened samples

reverse_transform(X)[source]

Unwhiten some whitened tensor X using the fitted PCA model.

Parameters:

Xtensor of shape (batch_size, self.n_eigenvec)

whitened input tensor

Returns:

unwhitened_Xtensor of shape (batch_size, vocabulary_size)

Batch of unwhitened centered samples