preprocess_focusE_weights

ampligraph.utils.preprocess_focusE_weights(data, weights, normalize=True)

Preprocessing of focusE weights.

Extract weights from data, remove NaNs, average weights and normalize them if self.focusE_params['normalize_numeric_values']==True.

Parameters:

  • data (array-like, shape (n,m)) – Array of shape (n,m) with \(m=4\). If weights=None, data contains triples and weights (\(m>3\)). If weights is passed, data only contains triples (\(m=3\)).

  • weights (array-like) – If not None, weights has shape (n, m-3), with m>0.

  • normalize (bool) – Specify whether to normalize the weights into the [0,1] range (default: True).

Returns:

processed_weights – An array of weights properly preprocessed and averaged into a unique vector if more than one vector of weights were given.

Return type:

np.array, shape (n, 1)