co_occurrence.py 文件源码

def vector_from_array(array):
    """Get triangle of output in vector from a correlation-type array

    Parameters
    ----------
    array : np.array

    Returns
    -------
    vector (np.array)

    Notes
    -----
    Old Matlab code indexes by column (aka.Fortran-style), so to get the indices
    of the top triangle, we have to do some reshaping.
    Otherwise, if the vector made up by rows is OK, then simply :
    triangle = np.triu_indices(array.size, k=1), out = array[triangle]
    """
    triangle_lower = np.tril_indices(array.shape[0], k=-1)
    flatten_idx = np.arange(array.size).reshape(array.shape)[triangle_lower]
    triangle = np.unravel_index(flatten_idx, array.shape, order='F')

    # triangle = np.triu_indices(array.size, k=1)
    # out = array[triangle]

    return array[triangle]