nif.data

class nif.data.PointWiseData(parameter_data, x_data, u_data, sample_weight=None)

Bases: object

Represents point-wise data.

Variables:

  • data_raw (numpy.ndarray) – Raw data.

  • data (numpy.ndarray) – Normalized data.

  • sample_weight (numpy.ndarray) – Sample weights.

  • n_p (int) – Number of parameter features.

  • n_x (int) – Number of state features.

  • n_o (int) – Number of output features.

property parameter

Returns the parameter data.

property x

Returns the state data.

property u

Returns the output data.

static standard_normalize(raw_data, area_weighted=False)

Performs standard normalization on raw data.

Parameters:

  • raw_data (numpy.ndarray) – Raw data.

  • area_weighted (bool) – Whether to perform area weighting. Defaults to False.

Returns:

numpy.ndarray – Normalized data. numpy.ndarray: Mean of raw data. numpy.ndarray: Standard deviation of raw data. numpy.ndarray: Normalized sample weights.

static minmax_normalize(raw_data, n_para, n_x, n_target, area_weighted=False)

Performs min-max normalization on raw data.

Parameters:

  • raw_data (numpy.ndarray) – Raw data.

  • n_para (int) – Number of parameter features.

  • n_x (int) – Number of state features.

  • n_target (int) – Number of output features.

  • area_weighted (bool) – Whether to perform area weighting. Defaults to False.

Returns:

numpy.ndarray – Normalized data. numpy.ndarray: Mean of raw data. numpy.ndarray: Standard deviation of raw data.

class nif.data.TFRDataset(n_feature, n_target, area_weight=False)

Bases: object

A class to handle creating and loading Tensorflow record datasets.

Parameters:

  • n_feature (int) – The number of features.

  • n_target (int) – The number of targets.

  • area_weight (bool, optional) – Whether or not to use area weights. Defaults to False.

create_from_npz(num_pts_per_file, npz_path, npz_key, tfr_path, prefix)

Create Tensorflow record files from a numpy file.

Parameters:

  • num_pts_per_file (int) – The number of points to put into each Tensorflow record file.

  • npz_path (str) – The path to the numpy file.

  • npz_key (str) – The key of the numpy array to use.

  • tfr_path (str) – The path to the output directory for the Tensorflow record files.

  • prefix (str) – The prefix to add to each Tensorflow record file name.

gen_dataset_from_batch_file(batch_file, batch_size)

Generate a TensorFlow Dataset from a batch file.

Parameters:

  • batch_file (np.ndarray) – A NumPy array containing the batch data.

  • batch_size (int) – The batch size.

Returns:

tf.data.Dataset – A TensorFlow Dataset object.

get_tfr_meta_dataset(tfr_path, epoch, tfr_shuffle_buffer_size=1)

Get a meta TensorFlow Dataset object from a folder of TFRecord files.

Parameters:

  • tfr_path (str) – The path to the folder containing the TFRecord files.

  • epoch (int) – The number of epochs to iterate through.

  • tfr_shuffle_buffer_size (int) – The shuffle buffer size.

Returns:

tf.data.Dataset – A TensorFlow Dataset object.