reconstuction errors¶

path orion.primitives.timeseries_errors.reconstruction_errors

description this primitive computes an array of errors comparing reconstructed and expected output. There are three main approaches for computing the discrepancies: point-wise, area, and dtw.

see json.

argument	type	description
parameters
`y`	`numpy.ndarray`	ground truth
`y_hat`	`numpy.ndarray`	predicted values
hyperparameters
`step_size`	`int`	indicates the number of steps between windows in the predicted values
`score_window`	`int`	indicates the size of the window over which the scores are calculated
`rec_error_type`	`str`	reconstruction error types, can be one of `["point", "area", "dtw"]`
`smooth`	`bool`	indicates whether the returned errors should be smoothed
`smoothing_window`	`float`	size of the smoothing window, expressed as a proportion of the total
output
`errors`	`numpy.ndarray`	array of errors

In [1]: import numpy as np

In [2]: from mlstars import load_primitive

In [3]: primitive = load_primitive('orion.primitives.timeseries_errors.reconstruction_errors')

In [4]: y = np.array([[1]] * 100)

In [5]: y_hat = np.array([[.99]] * 100)

In [6]: errors, predictions = primitive.produce(y=y, y_hat=y_hat)

In [7]: print("average error value: {:.2f}".format(errors.mean()))
average error value: 0.01

score anomalies regression errors