def likelihood_fit(time_diff,
model,
x0=None,
bounds=None,
title='',
legend='',
fit_min=-numpy.inf,
range_hist=(0., 20),
n_bins=100):
plt.subplot(2, 1, 1)
bins_center, n, dn = plot_functions.plot_histogram(time_diff,
"time [$\mu$s]",
"",
n_bins=n_bins,
range=range_hist,
title=title,
legend=legend,
fmt='.b',
as_scatter=True)
mask = bins_center > fit_min
bins_center = bins_center[mask]
n = n[mask]
minus_two_ll = functions.poisson_log_likelihood(bins_center, n, model)
res = minimize(minus_two_ll, x0=x0, bounds=bounds)
opt = res.x
print('OPT_likelihood 2expo:', opt)
plot_functions.scatter_plot(bins_center,
model(bins_center, *opt),
"time [$\mu$s]",
"",
fmt='-')
plt.subplot(2, 1, 2)
residuals = n - model(bins_center, *opt)
plot_functions.scatter_plot(bins_center,
residuals,
"time [$\mu$s]",
"res",
fmt='.')
return
def plot_channel_histogram(time_diff,
channel_start,
channel_stop,
n_bins,
fit_function=None,
param_names=None,
param_units=None,
p0=None,
bounds=(-numpy.inf, numpy.inf),
save_fig=False,
range_hist=None,
x_min=0.,
label='',
ex_int=(numpy.inf, -numpy.inf),
title=''):
print('\n')
legend = '%s%d eventi' % (label, len(time_diff))
bin_width = int(1000 * (range_hist[1] - range_hist[0]) / n_bins)
ylabel = 'ev./%d ns ' % (bin_width)
if fit_function is not None:
plt.subplot(2, 1, 1)
bins, n, dn = plot_functions.plot_histogram(time_diff,
"$\Delta t$ [$\mu$s]",
ylabel,
n_bins=n_bins,
range=range_hist,
title=title,
legend=legend,
fmt='.b',
as_scatter=True)
opt, pcov = plot_functions.do_fit(bins,
n,
dn,
param_names,
param_units,
fit_function=fit_function,
p0=p0,
bounds=bounds,
x_min=x_min,
ex_int=ex_int)
l_likelihood = functions.gauss_log_likelihood(bins, n, dn,
fit_function, *opt)
plt.subplot(2, 1, 2)
residuals = n - fit_function(bins, *opt)
plot_functions.scatter_plot(bins,
residuals,
"$\Delta t$ [$\mu$s]",
'Res.',
dx=None,
dy=dn,
title='')
#if save_fig == True:
# figlabel = 'dt_%d_%d_%s.pdf' % (channel_start, channel_stop, label)
# plt.savefig('%s' % figlabel , format = 'pdf')
return l_likelihood
else:
bins, n, dn = plot_functions.plot_histogram(time_diff,
"$\Delta t$ [$\mu$s]",
ylabel,
n_bins=n_bins,
range=range_hist,
title=title,
legend=legend,
fmt='.b',
as_scatter=True)
return bins, n, dn
dn,
param_names_2exp,
param_units_2exp,
fit_function=functions.two_expo,
p0=p0,
bounds=bounds,
x_min=fit_min,
x_max=x_max,
ex_int=ex_int)
#opt_expo, pcov_expo = plot_functions.do_fit(bins, n, dn, param_names = param_names, param_units = param_units, fit_function = functions.exponential, p0 = None, bounds = (-numpy.inf, numpy.inf), x_min = x_min, x_max = x_max)
plt.subplot(2, 1, 2)
residuals_two_expo = n - functions.two_expo(bins, *opt_two_expo)
plot_functions.scatter_plot(bins,
residuals_two_expo,
'time [$\mu$s]',
'residuals',
dx=None,
dy=dn,
title='')
#residuals_expo = n - functions.exponential(bins, *opt_expo)
#plot_functions.scatter_plot(bins, residuals_expo, 'time [$\mu$s]', 'residuals', dx = None, dy = dn/residuals_expo, title = '')
"""
#ALCUNI PLOT DI MONITORAGGIO
plt.figure()
n_bins = 20
plt.subplot(2, 3, 1)
index, channel_diff, time_diff = utilities.mask_array(ch, time, ch_stop_down, ch_stop_up)
plot_channel_histogram(time_diff, ch_stop_down, ch_stop_up, n_bins = n_bins, range_hist = (-0.1, 0.1), save_fig=save_fig)
plt.subplot(2, 3, 2)
index, channel_diff, time_diff = utilities.mask_array(ch, time, ch_stop_down, ch_stop_down)
plot_channel_histogram(time_diff, ch_stop_down, ch_stop_down, n_bins = n_bins, save_fig=save_fig)
# repeat experiments and average running time
for num_outlier in num_outliers:
running_times[num_outlier] = []
for t in range(repeat):
# score in 2d
rank_matrix = []
construction_time, scoring_time = 0., 0.
for i, features in enumerate(feature_pairs):
# Generate Plot
Y = features[0]
X = features[1]
rank_list, con_time, scor_time = plot_functions.scatter_plot(
eval(X), eval(Y), IDs, discription[Y], discription[X],
discription[Y] + ' vs ' + discription[X],
compare_value[X]) # score all points in 2d
rank_matrix.append(rank_list)
construction_time += con_time
scoring_time += scor_time
features = combine_features([
eval(F) for F in identity_features + continuous_features +
discrete_features
])
iForest(features)
start_time = time.time()
scaled_matrix, normal_matrix = ranklist.generate_graph(
P_val, num_outlier, rank_matrix)
# time_graph_gen = (time.time() - start_time) * num_outlier / (
epsilon_d3, epsilon_err_d3 = efficiency(N_sel, N_tot)
#Misure 18 maggio
N_sel = numpy.array([2502, 2140, 2904, 2083, 2114, 2450])
N_tot = numpy.array([9103, 2210, 3211, 2200, 2208, 3636])
epsilon_d4, epsilon_err_d4 = efficiency(N_sel, N_tot)
days = numpy.array([1., 2., 3., 4.])
epsilon_matrix = numpy.vstack([epsilon_d1, epsilon_d2, epsilon_d3, epsilon_d4])
epsilon_err_matrix = numpy.vstack(
[epsilon_err_d1, epsilon_err_d2, epsilon_err_d3, epsilon_err_d4])
plt.figure()
for i in range(len(epsilon_d1)):
legend = '%d' % (i + 1)
plot_functions.scatter_plot(days,
epsilon_matrix[:, i] / 100,
'days',
'epsilon',
dx=None,
dy=epsilon_err_matrix[:, i] / 100,
title='',
fmt='o-',
legend=legend)
epsilon_mean = numpy.mean(epsilon_matrix[:, i] / 100)
#epsilon_stdev =
print("epsilon_mean: ", epsilon_mean)
plt.ion()
plt.show()
"",
n_bins=n_bins,
range=range_hist,
title=title,
legend=legend,
fmt='.b',
as_scatter=True)
#ASIMMETRIA
asimmetry, asimmetry_err, bins_center = calculate_asimmetry(
time_diff_up, time_diff_down, n_bins, range_hist, x_min)
plt.figure()
plt.subplot(3, 1, 1)
plot_functions.scatter_plot(bins_center,
asimmetry,
'dt [$\mu$s]',
'Asimmetry ',
dy=asimmetry_err,
title='')
p0 = [0.1, 3., 0.0, 0.1]
bounds = (0., 0., -numpy.inf, -numpy.inf), (0.3, numpy.inf, 2 * numpy.pi,
+numpy.inf)
param_names = ['Amplitude', '$\omega$', '$\phi$', 'costant']
param_units = ['', 'MHz', 'rad', '']
opt_wave, pcov_wave = plot_functions.do_fit(bins_center,
asimmetry,
asimmetry_err,
param_names,
param_units,
functions.wave,
p0=p0,
bounds=bounds,
def likelihood_fit(model,
param_names,
param_units,
time_diff=None,
bin_center=None,
n=None,
jacobian=None,
x0=None,
bounds=None,
title='',
legend='',
fit_min=-numpy.inf,
range_hist=(0., 20),
n_bins=100,
output_file=''):
plt.subplot(3, 1, (1, 2))
if time_diff is not None:
bin_width = 100 #int(1000 * (range_hist[1]- range_hist[0])/n_bins)
ylabel = 'ev./%d ns ' % (bin_width)
bin_center, n, dn = plot_functions.plot_histogram(
time_diff,
"$\Delta t$ [$\mu$s]",
ylabel,
n_bins=n_bins,
range=range_hist,
title=title,
legend=legend,
fmt='.b',
as_scatter=True)
else:
bin_width = int(1000 * (bin_center[2] - bin_center[1]))
ylabel = 'ev./%d ns' % (bin_width)
dn = numpy.sqrt(n)
plot_functions.scatter_plot(bin_center,
n,
"$\Delta t$ [$\mu$s]",
ylabel,
dy=dn,
title=title,
fmt='.b')
mask = bin_center > fit_min
bin_center = bin_center[mask]
n = n[mask]
dn = dn[mask]
minus_two_ll = functions.poisson_log_likelihood(bin_center, n, model)
if jacobian is not None:
jac = jacobian(bin_center, n)
else:
jac = None
res = minimize(minus_two_ll,
x0=x0,
bounds=bounds,
method='BFGS',
jac=jac,
options={'disp': True})
opt = res.x
pcov = res.hess_inv
L = minus_two_ll(opt)
param_results = plot_functions.fit_legend(opt, numpy.sqrt(pcov.diagonal()),
param_names, param_units)
plot_functions.scatter_plot(bin_center,
model(bin_center, *opt),
"$\Delta t$ [$\mu$s]",
ylabel,
fmt='-',
legend=param_results,
title=title)
plt.subplot(3, 1, 3)
residuals = n - model(bin_center, *opt)
plot_functions.scatter_plot(bin_center,
residuals,
"$\Delta t$ [$\mu$s]",
"res",
fmt='.')
if output_file is not '':
param_results = param_results + '\n\n'
with open(output_file, 'a') as of:
of.write(param_results)
return L, bin_center, n, dn
TOF_true,
res=None)
T23, res23 = signal_propagation_functions.DT_23(x1[mask],
x3[mask],
y3[mask],
delay_T23,
TOF_true,
res=None)
plot_functions.multiple_histogram(x1[mask],
x3[mask],
"x1[mask]",
"x3[mask]",
bins=45)
plot_functions.multiple_histogram(T13, T23, "T13", "T23", bins=45)
plot_functions.scatter_plot(T12, T13, "T12 [ns]", "T13 [ns]")
plot_functions.scatter_plot(T23, T13, "T23 [ns]", "T13 [ns]")
plot_functions.histogram(TOF_true,
"TOF_true [ns]",
"dN/dTOF",
bins=100,
range=(6., 9.),
f=False)
plot_functions.hist2d(T23,
T13,
"T23",
"T13",
range_x=(15., 40.),
range_y=(15., 40.),
norm=LogNorm())
# get global list of outliers
features = combine_features([eval(F) for F in
identity_features + continuous_features + discrete_features])
iForest(features)
feature_pairs = generate_pairs(continuous_features,
continuous_features + discrete_features)
rank_matrix = []
for i, features in enumerate(feature_pairs):
# Generate Plot
Y = features[0]
X = features[1]
rank_list = plot_functions.scatter_plot(
eval(X), eval(Y), IDs, discription[Y], discription[X],
discription[Y] + ' vs ' + discription[X], compare_value[X]
)
rank_matrix.append(rank_list)
# select plots
scaled_matrix, normal_matrix = ranklist.generate_graph(P_val,
num_outlier, rank_matrix)
plots = plotSpot(budget, scaled_matrix, 'SpellOut')
# frequencies = generate_frequency_list(plots, scaled_matrix)
# end clock
time_elapsed = time.time() - start_time
running_times[num_edge].append(time_elapsed)
print num_edge, t, time_elapsed
import functions import utilities N_1 = numpy.array([29592/291, 23514/60, 27370/377]) N_2 = numpy.array([383850/385, 299636/324, 221795/306]) N_3 = numpy.array([39937/152, 42673/120, 44482/166]) N_4 = numpy.array([701930/151, 615246/120, 527184/100, 62384/100, 98885/126]) N_5 = numpy.array([38041/151, 31289/122, 29685/130]) N_6 = numpy.array([18258/151, 14259/120, 18306/153]) dN_1 = numpy.sqrt(N_1) dN_2 = numpy.sqrt(N_2) dN_3 = numpy.sqrt(N_3) dN_4 = numpy.sqrt(N_4) dN_5 = numpy.sqrt(N_5) dN_6 = numpy.sqrt(N_6) days = numpy.array([1., 2., 3.]) days_4 = numpy.array([1., 2., 2.5, 2.6, 3.]) plt.figure() plot_functions.scatter_plot(days, N_1, 'days', ' Noise[Hz]', dx = None, dy = dN_1, title = '', fmt='o-', legend = 'S1') plot_functions.scatter_plot(days, N_2, 'days', ' Noise[Hz]', dx = None, dy = dN_2, title = '', fmt='o-', legend = 'S2') plot_functions.scatter_plot(days, N_3, 'days', ' Noise[Hz]', dx = None, dy = dN_3, title = '', fmt='o-', legend = 'S3') plot_functions.scatter_plot(days_4, N_4, 'days', ' Noise[Hz]', dx = None, dy = dN_4, title = '', fmt='o-', legend = 'S4') plot_functions.scatter_plot(days, N_5, 'days', ' Noise[Hz]', dx = None, dy = dN_5, title = '', fmt='o-', legend = 'S5') plot_functions.scatter_plot(days, N_6, 'days', ' Noise[Hz]', dx = None, dy = dN_6, title = '', fmt='o-', legend = 'S6') plt.ion() plt.show()