This API reference is not for the latest stable Salvus version.

`salvus.opt.misfits.graph_space_optimal_transport`

Functions

`graph_space_optimal_transport_misfit_and_adoint_source()`

def graph_space_optimal_transport_misfit_and_adoint_source(
    data_synthetic: numpy.ndarray,
    data_observed: numpy.ndarray,
    sampling_rate_in_hertz: float,
    p: Union[float, int],
    max_expected_time_shift: Union[
        int, numpy.int32, numpy.int64, float, numpy.float32, numpy.float64
    ],
    plot: bool = False,
    return_assignment: bool = False,
) -> Union[
    Tuple[float, numpy.ndarray],
    Tuple[float, numpy.ndarray, numpy.ndarray, numpy.ndarray],
]:
    ...

Compute a graph-space optimal transport based misfit and adjoint source.

It largely implements it as described here https://doi.org/10.1088/1361-6420/ab206f but uses a different algorithm to solve the assignment problem.

Parameters

data_synthetic numpy.ndarray — synthetic waveforms.
data_observed numpy.ndarray — observed waveforms.
sampling_rate_in_hertz float — Sampling rate.
p Union[float, int] — Minkowski p-norm to use for the distance computation in the graph/point-cloud space.
max_expected_time_shift Union[int, numpy.int32, numpy.int64, float, numpy.float32, numpy.float64] — The maximum expected time shift in seconds.
plot bool — Plot the transport paths and the adjoint source. Very useful for debugging and tuning.
return_assignment bool — Optionally return the distance matrix and the optimal assignment, which is useful for visualizing the optimal transport map.

Returns Union[Tuple[float, numpy.ndarray], Tuple[float, numpy.ndarray, numpy.ndarray, numpy.ndarray]] — tuple of misfit value and adjoint source by default, or a tuple of mifit value, adjoint source, distance matrix and the assignment when return_assignment is True.