pysiglib.SigWindowStream

Added in version v3.0.0.

class SigWindowStream(dimension, degree, window_size, *, stride=1, scalar_term=False, n_jobs=1, _sig=None)

A fixed-width sliding window over a stream of incoming points that emits windowed signatures. A new window is emitted every stride points once enough points have been accumulated to fill the window.

Accepts a single path or a batch of independent paths - the batch shape is inferred from the first push / push_batch call. Windows are extracted along the time axis only; each batch element is windowed independently.

Parameters:

• dimension (int) – Dimension of the underlying space, \(d\).

• degree (int) – Truncation level of the signature, \(N\).

• window_size (int) – Number of points per window.

• stride (int) – Number of points between successive window starts. Default 1.

• scalar_term (bool) – If True, each emitted window signature includes the leading constant 1. If False (default), the leading element is stripped.

• n_jobs (int) – Number of threads to run in parallel in the internal per-window sig calls. -1 uses all available threads.

Example:

import pysiglib
import numpy as np

ws = pysiglib.SigWindowStream(dimension=3, degree=4, window_size=20, stride=5)

# Single path
path = np.random.randn(100, 3)
ws.push_batch(path)
window_sigs = ws.sig()  # shape (num_windows, sig_length)

# Batch of 8 independent paths
ws = pysiglib.SigWindowStream(dimension=3, degree=4, window_size=20, stride=5)
paths = np.random.randn(8, 100, 3)
ws.push_batch(paths)
window_sigs = ws.sig()  # shape (num_windows, 8, sig_length)

Citation

If you found this library useful in your research, please consider citing the paper:

@article{shmelev2025pysiglib,
  title={pySigLib-Fast Signature-Based Computations on CPU and GPU},
  author={Shmelev, Daniil and Salvi, Cristopher},
  journal={arXiv preprint arXiv:2509.10613},
  year={2025}
}