Function stump

tests/naive.py:225–344

Traverse distance matrix diagonally and update the top-k matrix profile and matrix profile indices if the parameter `row_wise` is set to `False`. If the parameter `row_wise` is set to `True`, it is a row-wise traversal.

(
    T_A,
    m,
    T_B=None,
    exclusion_zone=None,
    row_wise=False,
    k=1,
    T_A_subseq_isconstant=None,
    T_B_subseq_isconstant=None,
)

def stump(
    T_A,
    m,
    T_B=None,
    exclusion_zone=None,
    row_wise=False,
    k=1,
    T_A_subseq_isconstant=None,
    T_B_subseq_isconstant=None,
):
    """
    Traverse distance matrix diagonally and update the top-k matrix profile and
    matrix profile indices if the parameter `row_wise` is set to `False`. If the
    parameter `row_wise` is set to `True`, it is a row-wise traversal.
    """
    if T_B is None:  # self-join:
        ignore_trivial = True
        distance_matrix = np.array(
            [distance_profile(Q, T_A, m) for Q in core.rolling_window(T_A, m)]
        )
        T_B = T_A.copy()
        T_B_subseq_isconstant = T_A_subseq_isconstant
    else:
        ignore_trivial = False
        distance_matrix = np.array(
            [distance_profile(Q, T_B, m) for Q in core.rolling_window(T_A, m)]
        )

    T_A_subseq_isconstant = rolling_isconstant(T_A, m, T_A_subseq_isconstant)
    T_B_subseq_isconstant = rolling_isconstant(T_B, m, T_B_subseq_isconstant)

    distance_matrix[np.isnan(distance_matrix)] = np.inf
    for i in range(distance_matrix.shape[0]):
        for j in range(distance_matrix.shape[1]):
            if np.isfinite(distance_matrix[i, j]):
                if T_A_subseq_isconstant[i] and T_B_subseq_isconstant[j]:
                    distance_matrix[i, j] = 0.0
                elif T_A_subseq_isconstant[i] or T_B_subseq_isconstant[j]:
                    distance_matrix[i, j] = np.sqrt(m)
                else:  # pragma: no cover
                    pass

    n_A = T_A.shape[0]
    n_B = T_B.shape[0]
    l = n_A - m + 1
    if exclusion_zone is None:
        exclusion_zone = int(np.ceil(m / config.STUMPY_EXCL_ZONE_DENOM))

    P = np.full((l, k + 2), np.inf, dtype=np.float64)
    I = np.full((l, k + 2), -1, dtype=np.int64)  # two more columns are to store
    # ... left and right top-1 matrix profile indices

    if row_wise:  # row-wise traversal in distance matrix
        if ignore_trivial:  # self-join
            for i in range(l):
                apply_exclusion_zone(distance_matrix[i], i, exclusion_zone, np.inf)

        for i, D in enumerate(distance_matrix):  # D: distance profile