gw.training.train_builders.build_svd_for_embedding_network

gw.training.train_builders.build_svd_for_embedding_network(
    wfd,
    data_settings,
    asd_dataset_path,
    size,
    num_training_samples,
    num_validation_samples,
    num_workers=0,
    batch_size=1000,
    out_dir=None,
)

Construct SVD matrices V based on clean waveforms in each interferometer. These will be used to seed the weights of the initial projection part of the embedding network.

It first generates a number of training waveforms, and then produces the SVD.

Parameters

Name	Type	Description	Default
wfd	WaveformDataset		required
data_settings	dict		required
asd_dataset_path	str	Training waveforms will be whitened with respect to these ASDs.	required
size	int	Number of basis elements to include in the SVD projection.	required
num_training_samples	int		required
num_validation_samples	int		required
num_workers	int		0
batch_size	int		1000
out_dir	str	SVD performance diagnostics are saved here.	None

Returns

Name	Type	Description
	list of numpy arrays	The V matrices for each interferometer. They are ordered as in data_settings[ ‘detectors’].