pyhgf.utils.vectorised_belief_propagation.batched_prediction_pass#

pyhgf.utils.vectorised_belief_propagation.batched_prediction_pass(network, x)[source]#

Forward-only sweep for a batch of samples, compiled once and reused.

The batched equivalent of prediction_pass(): every row of x is an independent sample swept from the same network state. Used by pyhgf.model.DeepNetwork.predict() so repeated batched calls hit the compilation cache instead of rebuilding the batching wrapper.

Parameters:

  • network (VectorisedNetwork) – The current vectorised network state.

  • x (Array) – Predictors, shape (batch, n_input_features).

Returns:

The bottom element’s expected_mean per sample, shape (batch, n_output_features).

Return type:

expected_mean