def barrierScore(xd, yd, pp, weights=None, thr=3e9):
""" Calculate score for barrier model """
ppq = pp.copy()
# ppq[0]+=3.05
ydatax = barrierModel(xd, ppq)
sc = np.abs(ydatax - yd)
scalefac = thr
#sc=robustCost(sc, thr)
sc = np.sqrt(robustCost(sc / scalefac, thr / scalefac, 'BZ0')) * scalefac
if weights is not None:
sc = sc * weights
sc = np.linalg.norm(sc, ord=4) / sc.size
return sc
def pat_one_ele_score(self, xd, yd, pp, weights=None, thr=2e9):
""" Calculate score for pat one electron model
Args:
xd (array): x coordinates of peaks in sensor signal
yd (array): y coordinates of peaks in sensor signal
pp (array): model parameters
"""
ydatax = one_ele_pat_model(xd, pp)
charge_change = np.abs(np.abs(pp[1]) * (xd - pp[0]) / np.sqrt((pp[1] * (xd - pp[0]))**2 + 4 * pp[2]**2))
sc = np.abs(ydatax - yd) * charge_change
scalefac = thr
sc = np.sqrt(robustCost(sc / scalefac, thr / scalefac, 'BZ0')) * scalefac
if weights is not None:
sc = sc * weights
sc = np.linalg.norm(sc, ord=4) / sc.size
if pp[1] < 10:
sc *= 10
if pp[2] > 150:
sc *= 10
return sc
def pat_two_ele_score(self, xd, yd, pp, weights=None, thr=2e9):
""" Calculate score for pat two electron model
Args:
xd (array): x coordinates of peaks in sensor signal
yd (array): y coordinates of peaks in sensor signal
pp (array): model parameters
"""
ymodel = two_ele_pat_model(xd, pp)
denom = np.sqrt((pp[1] * (xd - pp[0]))**2 + 8 * pp[2]**2)
charge_changes = []
charge_changes.append(1 / 2 * (1 + pp[1] * (xd - pp[0]) / denom))
charge_changes.append(np.abs(pp[1] * (xd - pp[0]) / denom))
charge_changes.append(1 / 2 * (1 - pp[1] * (xd - pp[0]) / denom))
linesize = ymodel[0].shape[0]
if self.branch_reduction is None or self.branch_reduction == 'minimum':
sc = np.inf * np.ones(linesize)
for idval, val in enumerate(self.even_branches):
if val:
sc = np.minimum(sc, np.abs(ymodel[idval] - yd))
elif self.branch_reduction == 'mean':
sc = []
for idval, val in enumerate(self.even_branches):
if val:
sc.append(np.abs(ymodel[idval] - yd))
sc = np.mean(np.array(sc), axis=1)
else:
raise NotImplementedError('branch_reduction %s not implemented' %
self.branch_reduction)
scalefac = thr
sc = np.sqrt(robustCost(sc / scalefac, thr / scalefac,
'BZ0')) * scalefac
if weights is not None:
sc *= weights
sc = np.linalg.norm(sc, ord=4) / sc.size
if pp[1] < 10:
sc *= 10000
return sc
def test_robust_cost():
x = np.array([0, 1, 2, 3, 4, 5])
_ = pgeometry.robustCost(x, 2)
_ = pgeometry.robustCost(x, 'auto')
