ranking_HM.Rank.rank_scores_calc_MZ#

Back to Class page

Summary#

Calculating ranking scores for all the bg, fg, lsc, inj triggers. The rank functions are constructed in a sub-function make_score_funcs_MZ(). We first group the templates based on their “glitchiness” and then make separate rank functions for different groups

Signature#

def rank_scores_calc_MZ(self, i_subbank, safety_factor = 4, matching_point = None, downsampling_correction = True, min_trigs_per_grp = 500, output_rank_func_temp_grps = False, group_coarse_calphas = False, n_calpha_dim = 2, snr2_ref = None, n_snr2_ref = 0.001, vetoed = True, rhosqs_arrays = None, calphas_arrays = None, masks_vetoed = None, **score_func_kwargs)

Input variables#
Name	Type	Default	Description
`i_subbank`			index of the subbank of the bank
`safety_factor`		4	Add this to threshold_chi2 before estimating the rank functions to account for incompleteness related to optimization
`matching_point`		None	Set SNR^2 at which the rank functions are matched, the default is threshold network SNR^2/2 (the rank fns are made such that all template groups within the same subbank match at self.snr2min and the least glitchy template group in all subbanks match at matching_point)
`downsampling_correction`		True	If the triggers were downsampled compared to a chi-sq distribution because of an additional cut (e.g., based on whether the mode ratios A33/A22 or A44/A22 are physical). This flag corrects the rank function so that it follows the chi-sq behavior again. This flag needs a file downsamp_corr_path.npy to be input when creating Rank class object
`min_trigs_per_grp`		500	To avoid pathologies with making the rank functions, we require that the templates in each group have more than a particular number of background triggers associated to them
`output_rank_func_temp_grps`		False	Flag if you want to output the rank funcs for separate template groups for reference
`group_coarse_calphas`		False	Flag whether we group the coarse or fine calphas
`n_calpha_dim`		2	Number of dimensions to use for grouping calphas
`snr2_ref`		None	Quantify templates’ glitchiness by the fraction of triggers with SNR>snr2_ref
`n_snr2_ref`		0.001	Set snr2_ref by demanding that in the Gaussian noise case, the expected number of triggers > snr2_ref is n_snr2_ref over all calphas. Used only if snr2_ref is None
`vetoed`		True	Flag to compute scores only for vetoed triggers
`rhosqs_arrays`		None	If known, 4 x n_event x 2 array of SNR^2 values for background, foreground, LVC, injections (pass to avoid looping over hdf5)
`calphas_arrays`		None	If known, 4 x n_event x n_calpha_dim array of calphas
`masks_vetoed`		None	If known, array of length 4 with masks for vetoed triggers This makes us set vetoed to True
`\\score_func_kwargs`

Output variables#

Return annotation	Docstring type	Description
`None`

Docstring#

Calculating ranking scores for all the bg, fg, lsc, inj triggers.
The rank functions are constructed in a sub-function
make_score_funcs_MZ(). We first group the templates based on their
"glitchiness" and then make separate rank functions for different groups
:param i_subbank: index of the subbank of the bank
:param safety_factor:
    Add this to threshold_chi2 before estimating the rank functions
    to account for incompleteness related to optimization
:param matching_point:
    Set SNR^2 at which the rank functions are matched, the default
    is threshold network SNR^2/2
    (the rank fns are made such that all template groups within the same
     subbank match at self.snr2min and the least glitchy template group
     in all subbanks match at matching_point)
:param downsampling_correction:
    If the triggers were downsampled compared to a chi-sq distribution
    because of an additional cut (e.g., based on whether the mode ratios
    A33/A22 or A44/A22 are physical). This flag corrects the rank
    function so that it follows the chi-sq behavior again. This flag
    needs a file downsamp_corr_path.npy to be input when creating
    Rank class object
:param min_trigs_per_grp:
    To avoid pathologies with making the rank functions, we require that
    the templates in each group have more than a particular
    number of background triggers associated to them
:param output_rank_func_temp_grps:
    Flag if you want to output the rank funcs for separate template groups
    for reference
:param group_coarse_calphas:
    Flag whether we group the coarse or fine calphas
:param n_calpha_dim: Number of dimensions to use for grouping calphas
:param snr2_ref:
    Quantify templates' glitchiness by the fraction of triggers with
    SNR>snr2_ref
:param n_snr2_ref:
    Set snr2_ref by demanding that in the Gaussian noise case, the
    expected number of triggers > snr2_ref is n_snr2_ref over all
    calphas. Used only if snr2_ref is None
:param vetoed: Flag to compute scores only for vetoed triggers
:param rhosqs_arrays:
    If known, 4 x n_event x 2 array of SNR^2 values for background,
    foreground, LVC, injections (pass to avoid looping over hdf5)
:param calphas_arrays:
    If known, 4 x n_event x n_calpha_dim array of calphas
:param masks_vetoed:
    If known, array of length 4 with masks for vetoed triggers
    This makes us set vetoed to True