pysiglib.branched_sig_kernel_backprop

pysiglib.branched_sig_kernel_backprop#

Warning

Where possible, pysiglib.torch_api should be used rather than explicitly calling backpropagation functions. Explicit backpropagation can introduce subtle errors if called incorrectly.

branched_sig_kernel_backprop(derivs, path1, path2, depth, dyadic_order, *, static_kernel=None, time_aug=False, lead_lag=False, end_time=1.0, left_deriv=True, right_deriv=False, k_stack=None, n_jobs=1, return_grid=False)[source]#

Backpropagates through pysiglib.branched_sig_kernel().

Parameters:

  • derivs (numpy.ndarray | torch.Tensor) – Derivatives with respect to a scalar branched kernel or the final-depth grid when return_grid=True.

  • path1 (numpy.ndarray | torch.Tensor) – First path or batch of paths.

  • path2 (numpy.ndarray | torch.Tensor) – Second path or batch of paths.

  • depth (int) – Forest depth truncation used in the forward pass.

  • dyadic_order (int | tuple) – Dyadic refinement order, or a pair of orders.

  • k_stack (None | numpy.ndarray | torch.Tensor) – Optional internal all-depth grid stack. If omitted, it is reconstructed from the static-kernel increments.

  • return_grid (bool) – If True, derivs is interpreted as final-grid derivatives.

Returns:

Derivatives with respect to one or both paths.

Return type:

tuple


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}
}
Contents