Label

Functions to label synthetic data.

ecgan.utils.label.count_nonzero(input, dim=None) → Tensor

Counts the number of non-zero values in the tensor input along the given dim. If no dim is specified then all non-zeros in the tensor are counted.

Parameters

• input (Tensor) -- the input tensor.

• dim (int or tuple of ints, optional) -- Dim or tuple of dims along which to count non-zeros.

Example:

>>> x = torch.zeros(3,3)
>>> x[torch.randn(3,3) > 0.5] = 1
>>> x
tensor([[0., 1., 1.],
        [0., 0., 0.],
        [0., 0., 1.]])
>>> torch.count_nonzero(x)
tensor(3)
>>> torch.count_nonzero(x, dim=0)
tensor([0, 1, 2])

ecgan.utils.label.label_generated_data_pointwise(anomaly_scores, tau)

Labeling of generated data depending on some tau.

Parameters

• anomaly_scores (Tensor) -- Pointwise anomaly scores

• tau (float) -- Anomaly threshold.

Return type

Tensor

Returns

The labels of generated series/data points.

ecgan.utils.label.label_data_by_summation(anomaly_scores, tau, channelwise=True)

Calculate one label per channel (or series if channelwise=False).

Utilizes the sum of pointwise anomaly scores and checks if the average anomaly score is below tau.

Parameters

• anomaly_scores (Tensor) -- Pointwise anomaly scores.

• tau -- (Pointwise) anomaly threshold.

• channelwise (bool) -- Flag to indicate if the data should be labeled channelwise.

Return type

Tensor

Returns

One label for each series or channel, meaning anomaly_scores.shape[0] labels for serieswise scoring and respectively anomaly_score.shape[0] * anomaly_score.shape[2] labels for channelwise detection will be returned.

ecgan.utils.label.label_data_by_variance(anomaly_scores, tau, channelwise=True)

Calculate one label per channel (or series if channelwise=False).

Utilizes the variance of pointwise anomaly scores and checks if the anomaly score is below the given tau.

Parameters

• anomaly_scores (Tensor) -- Pointwise anomaly scores.

• tau (float) -- (Pointwise) anomaly threshold.

• channelwise (bool) -- Flag indicating if you want to return channelwise or serieswise anomaly scores.

Return type

Tensor

Returns

One label for each series or channel, meaning anomaly_scores.shape[0] labels for serieswise scoring and respectively anomaly_score.shape[0] * anomaly_score.shape[2] labels for channelwise detection will be returned.

ecgan.utils.label.label_absolute(anomaly_scores, tau=0.2, anomaly_lower_bound=None)

Label channels based on the absolute amount of pointwise anomalies.

A channel is labeled as anomalous if more than anomaly_lower_bound samples are labeled during the pointwise detection.

Return type

Tensor

ecgan.utils.label.label(anomaly_scores, strategy=LabelingStrategy.POINTWISE, tau=0.2)

Label synthetic data based on the respective anomaly scores.

Parameters

• anomaly_scores (Tensor) -- Series of pointwise anomaly scores.

• strategy (LabelingStrategy) -- Labeling strategy: either pointwise, channelwise or serieswise.

• tau (float) -- Anomaly threshold.

Return type

Tensor

Returns

Labels for each data point, channel or series. User has to ensure the correct format.