augmentation

polars_ts.clustering._augmentation

Time series augmentations for contrastive learning.

jitter(x, sigma=0.1)

Add Gaussian noise to each time step.

scaling(x, sigma=0.1)

Multiply by a random per-sample scale factor.

window_crop(x, crop_ratio=0.5)

Randomly crop a contiguous window and return it.

augment_pair(x, jitter_sigma=0.1, scale_sigma=0.1)

Generate two augmented views of the input batch.