"""Given the single value for the weight parameter balancing between impostors and spoofing attacks and a threshold, calculates the error rates and their relationship depending on the criteria (difference in case of 'eer', hter in case of 'hter' criteria)
"""Given the single value for the weight parameter balancing between
impostors and spoofing attacks and a threshold, calculates the error rates
and their relationship depending on the criteria (difference in case of
'eer', hter in case of 'hter' criteria)
Keyword parameters:
- data - the development data used to determine the threshold. List on 4 numpy.arrays containing: negatives (licit), positives (licit), negatives (spoof), positivies (spoof)
- weight - the weight parameter balancing between impostors and spoofing attacks
- data - the development data used to determine the threshold. List on 4
- weight - the weight parameter balancing between impostors and spoofing
attacks
- thres - the given threshold
- beta - the weight parameter balancing between real accesses and all the negative samples (impostors and spoofing attacks). Note that this parameter will be overriden and not considered if the selected criteria is 'hter'.
- beta - the weight parameter balancing between real accesses and all the
negative samples (impostors and spoofing attacks). Note that this
parameter will be overriden and not considered if the selected criteria
is 'hter'.
- criteria - 'eer', 'wer' or 'hter' criteria for decision threshold
"""Recursive search for the optimal threshold given a criteria. It evaluates the full range of thresholds at 100 points, and computes the one which optimizes the threshold. In the next search iteration, it examines the region around the point that optimizes the threshold. The procedure stops when the search range is smaller then 1e-10.
"""Recursive search for the optimal threshold given a criteria. It
evaluates the full range of thresholds at 100 points, and computes the one
which optimizes the threshold. In the next search iteration, it examines
the region around the point that optimizes the threshold. The procedure
stops when the search range is smaller then 1e-10.
Keyword arguments:
- data - the development data used to determine the threshold. List on 4 numpy.arrays containing: negatives (licit), positives (licit), negatives (spoof), positivies (spoof)
- data - the development data used to determine the threshold. List on 4
- weight - the weight parameter balancing between impostors and spoofing attacks
- beta - the weight parameter balancing between real accesses and all the negative samples (impostors and spoofing attacks). Note that methods called within this function will override this parameter and not considered if the selected criteria is 'hter'.
- criteria - the decision threshold criteria ('eer' for EER, 'wer' for Minimum WER or 'hter' for Minimum HTER criteria).
- weight - the weight parameter balancing between impostors and spoofing
attacks
- beta - the weight parameter balancing between real accesses and all the
negative samples (impostors and spoofing attacks). Note that methods called
within this function will override this parameter and not considered if the
selected criteria is 'hter'.
- criteria - the decision threshold criteria ('eer' for EER, 'wer' for
"""Calculates the threshold for achieving the given criteria between the FAR_w and the FRR, given the single value for the weight parameter balancing between impostors and spoofing attacks and a single value for the parameter beta balancing between the real accesses and the negatives (impostors and spoofing attacks)
"""Calculates the threshold for achieving the given criteria between the
FAR_w and the FRR, given the single value for the weight parameter
balancing between impostors and spoofing attacks and a single value for the
parameter beta balancing between the real accesses and the negatives
(impostors and spoofing attacks)
Keyword parameters:
- licit_neg - numpy.array of scores for the negatives (licit scenario)
- licit_pos - numpy.array of scores for the positives (licit scenario)
- spoof_neg - numpy.array of scores for the negatives (spoof scenario)
- spoof_pos - numpy.array of scores for the positives (spoof scenario)
- weight - the weight parameter balancing between impostors and spoofing attacks
- beta - the weight parameter balancing between real accesses and all the negative samples (impostors and spoofing attacks). Note that methods called within this function will override this parameter and not considered if the selected criteria is 'hter'.
- criteria - the decision threshold criteria ('eer' for EER, 'wer' for Minimum WER or 'hter' for Minimum HTER criteria).
- weight - the weight parameter balancing between impostors and spoofing
attacks
- beta - the weight parameter balancing between real accesses and all the
negative samples (impostors and spoofing attacks). Note that methods called
within this function will override this parameter and not considered if the
selected criteria is 'hter'.
- criteria - the decision threshold criteria ('eer' for EER, 'wer' for
"""Calculates the optimal thresholds for EPSC, for a range of the weight parameter balancing between impostors and spoofing attacks, and for a range of the beta parameter balancing between real accesses and all the negatives (impostors and spoofing attacks)
"""Calculates the optimal thresholds for EPSC, for a range of the weight
parameter balancing between impostors and spoofing attacks, and for a range
of the beta parameter balancing between real accesses and all the negatives
- spoof_pos - numpy.array of scores for the positives (spoof scenario)
- points - number of points to calculate EPSC
- criteria - the decision threshold criteria ('eer', 'wer' or 'hter')
- omega - the value of the parameter omega, balancing between impostors and spoofing attacks. If None, it is going to span the full range [0,1]. Otherwise, can be set to a fixed value or a list of values.
- beta - the value of the parameter beta, balancing between real accesses and all the negatives (zero-effort impostors and spoofing attacks). If None, it is going to span the full range [0,1]. Otherwise, can be set to a fixed value or a list of values.
- omega - the value of the parameter omega, balancing between impostors and
spoofing attacks. If None, it is going to span the full range [0,1].
Otherwise, can be set to a fixed value or a list of values.
- beta - the value of the parameter beta, balancing between real accesses
and all the negatives (zero-effort impostors and spoofing attacks). If
None, it is going to span the full range [0,1]. Otherwise, can be set to a
"""Calculates several error rates: FRR, FAR (zero-effort impostors), SFAR, FAR_w, HTER_w for a given value of w. It returns the calculated threshold as a last argument
"""Calculates several error rates: FRR, FAR (zero-effort impostors), SFAR,
FAR_w, HTER_w for a given value of w. It returns the calculated threshold
"""Calculates several error rates: FAR_w and WER_wb for the given weights (omega and beta) and thresholds (the thresholds need to be computed first using the method: epsc_thresholds() before passing to this method)
"""Calculates several error rates: FAR_w and WER_wb for the given weights
(omega and beta) and thresholds (the thresholds need to be computed first
using the method: epsc_thresholds() before passing to this method)
"""Calculates several error rates: FAR_w and HTER_w for the given weights (omega and beta) and thresholds (the thresholds need to be computed first using the method: epsc_thresholds() before passing to this method)
"""Calculates several error rates: FAR_w and HTER_w for the given weights
(omega and beta) and thresholds (the thresholds need to be computed first
using the method: epsc_thresholds() before passing to this method)