def get_interped_y(label, data, x, x_to): run_y = lfu.grab_mobj_by_name(label, data) run_interped = lgeo.scalars( label = 'interpolated best result - ' + label, scalars = lm.linear_interpolation( x.scalars, run_y.scalars, x_to.scalars, 'linear')) return run_interped
def get_single_well_data(self):
# THIS FUNCTION WILL JUST RETURN THE SELECTED WELLS PLOT DATA
od_dex = self.get_selected_well()
time_ = self.data.data[0]
od = self.data.data[od_dex]
dbl = self._doubling_times_[od_dex]
gro = self._growth_rates_[od_dex]
lo, hi = self._thresholds_[od_dex]
#od.scalars = od.scalars
y_low = lgeo.scalars('__skip__', [lo]*2, color = 'black')
# [self.OD_threshold_low]*2, color = 'black')
y_high = lgeo.scalars('__skip__', [hi]*2, color = 'black')
# [self.OD_threshold_high]*2, color = 'black')
x_stack = lgeo.scalars('__skip__',
[time_.scalars.min(), time_.scalars.max()])
# CALCULATE THE DOUBLING TIME AND ADD VERTICAL LINES FOR THE POINTS USED
# MAKE ALL THESE FLAT LINES DASHED OR SOMETHING
#pdb.set_trace()
data = lfu.data_container(
domains = [time_, x_stack, x_stack],
codomains = [od, y_low, y_high],
doubling = dbl, growth_rate = gro,
thresholds = (lo, hi))
return data
def initialize(self, *args, **kwargs): if not self.cooling_curve: self.final_iteration =\ self.fitted_criteria[0].max_iterations lam = -1.0 * np.log(self.max_temperature)/\ self.final_iteration cooling_domain = np.array(range(self.final_iteration)) cooling_codomain = self.max_temperature*np.exp( lam*cooling_domain) self.cooling_curve = lgeo.scalars( label = 'cooling curve', scalars = cooling_codomain) fit_routine.initialize(self, *args, **kwargs) self.data.extend(lgeo.scalars_from_labels( ['annealing temperature'])) self.temperature = self.cooling_curve.scalars[self.iteration] self.parameter_space.initial_factor = self.temperature
def finalize(self, *args, **kwargs):
def get_interped_y(label, data, x, x_to):
run_y = lfu.grab_mobj_by_name(label, data)
run_interped = lgeo.scalars(
label = 'interpolated best result - ' + label,
scalars = lm.linear_interpolation(
x.scalars, run_y.scalars,
x_to.scalars, 'linear'))
return run_interped
self.best_fits = [(met.best_measure,
met.data[0].scalars[met.best_measure])
for met in self.metrics]
self.handle_fitting_key()
self.parameter_space.set_current_position(
self.p_sp_trajectory[self.best_fits[
self.prime_metric][0]])
best_run_data = lis.run_system(self.ensemble)
best_run_data = [lgeo.scalars(label = lab, scalars = dat)
for lab, dat in zip(self.run_targets, best_run_data)]
best_run_data_x = lfu.grab_mobj_by_name(
self.data_to_fit_to[0].label, best_run_data)
best_run_data_ys = [get_interped_y(
lab, best_run_data, best_run_data_x, self.data_to_fit_to[0])
for lab in lfu.grab_mobj_names(self.data_to_fit_to[1:])]
print 'fit routine:', self.label, 'best fit:', self.best_fits
print '\tran using regime:', self.regime
best_data = self.data_to_fit_to + best_run_data_ys
lgd.quick_plot_display(best_data[0], best_data[1:], delay = 5)
self.ensemble.data_pool.pool_names =\
[dat.label for dat in best_data]
self.ensemble.data_pool.batch_pool.append(best_data)
if self.regime.startswith('coarse'):
self.impose_coarse_result_to_p_space()
self.ensemble.cartographer_plan.parameter_space.\
set_current_position(self.p_sp_trajectory[
self.best_fits[self.prime_metric][0]])
def on_write_data(self, *args, **kwargs):
if not self.domain:
self.domain = ["x"]
self.codomains = ["y1", "y2", "y3"]
self.domain_data = np.linspace(0, 10, 100)
self.codomain_datas = [
np.exp(self.domain_data) / np.exp(self.domain_data[-1]),
np.sin(self.domain_data),
np.cos(2 * self.domain_data),
]
print "USED FAKE TEST DATA!!"
# the data should be normalized -1, 1 on both axes
x = self.domain_data
ys = self.codomain_datas
[plt.plot(x, y) for y in ys]
plt.show()
self.out_data = lfu.data_container(
data=[
lgeo.scalars(label=lab, scalars=dat)
for lab, dat in zip(self.domain + self.codomains, [self.domain_data] + self.codomain_datas)
]
)
lf.save_pkl_object(self.out_data, self.out_filename)
def on_write_data(self, *args, **kwargs):
if not self.domain:
self.domain = ['x']
self.codomains = ['y1', 'y2', 'y3']
self.domain_data = np.linspace(0, 10, 100)
self.codomain_datas = [
np.exp(self.domain_data) / np.exp(self.domain_data[-1]),
np.sin(self.domain_data),
np.cos(2 * self.domain_data)
]
print 'USED FAKE TEST DATA!!'
#the data should be normalized -1, 1 on both axes
x = self.domain_data
ys = self.codomain_datas
[plt.plot(x, y) for y in ys]
plt.show()
self.out_data = lfu.data_container(data=[
lgeo.scalars(label=lab, scalars=dat)
for lab, dat in zip(self.domain +
self.codomains, [self.domain_data] +
self.codomain_datas)
])
lf.save_pkl_object(self.out_data, self.out_filename)
