def predict_values(self, refresh=False):
self.debug('predict values {}'.format(refresh))
try:
x, y, z, ze, j, je, sj, sje = self._extract_position_arrays()
t = AffineTransform()
t.rotate(self.rotation)
x, y = t.transforms(x, y)
# print(x)
except ValueError as e:
self.debug('no monitor positions to fit, {}'.format(e))
return
# print(x)
# print(y)
# print(z)
# print(ze)
n = x.shape[0]
if n >= 3 or self.plotter_options.model_kind in (WEIGHTED_MEAN,
MATCHING, BRACKETING):
# n = z.shape[0] * 10
r = max((max(abs(x)), max(abs(y))))
# r *= 1.25
reg = self._regressor_factory(x, y, z, ze)
self._regressor = reg
else:
msg = 'Not enough monitor positions. At least 3 required. Currently only {} active'.format(
n)
self.debug(msg)
self.information_dialog(msg)
return
options = self.plotter_options
ipositions = self.unknown_positions + self.monitor_positions
if options.model_kind == LEAST_SQUARES_1D:
k = options.one_d_axis.lower()
pts = array([getattr(p, k) for p in ipositions])
else:
pts = array([[p.x, p.y] for p in ipositions])
if options.use_monte_carlo and options.model_kind not in (MATCHING,
BRACKETING,
NN):
fe = FluxEstimator(options.monte_carlo_ntrials, reg)
split = len(self.unknown_positions)
nominals, errors = fe.estimate(pts)
if options.position_error:
_, pos_errors = fe.estimate_position_err(
pts, options.position_error)
else:
pos_errors = zeros(pts.shape[0])
for positions, s, e in ((self.unknown_positions, 0, split),
(self.monitor_positions, split, None)):
noms, es, ps = nominals[s:e], errors[s:e], pos_errors[s:e]
for p, j, je, pe in zip(positions, noms, es, ps):
oj = p.saved_j
p.j = j
p.jerr = je
p.position_jerr = pe
p.dev = (oj - j) / j * 100
else:
js = reg.predict(pts)
jes = reg.predict_error(pts)
for j, je, p in zip(js, jes, ipositions):
p.j = float(j)
p.jerr = float(je)
p.dev = (p.saved_j - j) / j * 100
p.mean_dev = (p.mean_j - j) / j * 100
if options.plot_kind == '2D':
self._graph_contour(x, y, z, r, reg, refresh)
elif options.plot_kind == 'Grid':
self._graph_grid(x, y, z, ze, r, reg, refresh)
else:
if options.model_kind in (LEAST_SQUARES_1D, WEIGHTED_MEAN_1D):
self._graph_linear_j(x, y, r, reg, refresh)
else:
self._graph_hole_vs_j(x, y, r, reg, refresh)
def predict_values(self, refresh=False):
self.debug('predict values {}'.format(refresh))
try:
x, y, z, ze, j, je, sj, sje = self._extract_position_arrays()
t = AffineTransform()
t.rotate(self.rotation)
x, y = t.transforms(x, y)
# print(x)
except ValueError as e:
self.debug('no monitor positions to fit, {}'.format(e))
return
# print(x)
# print(y)
# print(z)
# print(ze)
n = x.shape[0]
if n >= 3:
# n = z.shape[0] * 10
r = max((max(abs(x)), max(abs(y))))
# r *= 1.25
reg = self._regressor_factory(x, y, z, ze)
self._regressor = reg
else:
msg = 'Not enough monitor positions. At least 3 required. Currently only {} active'.format(n)
self.debug(msg)
self.information_dialog(msg)
return
options = self.plotter_options
ipositions = self.unknown_positions + self.monitor_positions
pts = array([[p.x, p.y] for p in ipositions])
if options.use_monte_carlo and options.model_kind not in ['Matching', 'Bracketing']:
fe = FluxEstimator(options.monte_carlo_ntrials, reg)
split = len(self.unknown_positions)
nominals, errors = fe.estimate(pts)
if options.position_error:
_, pos_errors = fe.estimate_position_err(pts, options.position_error)
else:
pos_errors = zeros(pts.shape[0])
for positions, s, e in ((self.unknown_positions, 0, split),
(self.monitor_positions, split, None)):
noms, es, ps = nominals[s:e], errors[s:e], pos_errors[s:e]
for p, j, je, pe in zip(positions, noms, es, ps):
oj = p.saved_j
p.j = j
p.jerr = je
p.position_jerr = pe
p.dev = (oj - j) / j * 100
else:
js = reg.predict(pts)
jes = reg.predict_error(pts)
for j, je, p in zip(js, jes, ipositions):
p.j = float(j)
p.jerr = float(je)
p.dev = (p.saved_j - j) / j * 100
if options.plot_kind == '2D':
self._graph_contour(x, y, z, r, reg, refresh)
elif options.plot_kind == 'Grid':
self._graph_grid(x, y, z, ze, r, reg, refresh)
else:
self._graph_hole_vs_j(x, y, r, reg, refresh)
