def main():
solar_mass = 2. * 10**30
m1 = 1.4 * solar_mass
m2 = 1.4 * solar_mass
mc = (m1 * m2)**(3. / 5.) / (m1 + m2)**(1. / 5.)
r = 1000.
t_coal = 1
phi_coal = 0
iota = np.pi / 8
theta = 0
phi = 0
psi = 0
t_dur = 0.15
t_max = t_coal - (t_dur / 100.)
t_from_merger = 0
#print("M_c: {}".format(mc))
num_pts = 10000
x = np.linspace(t_max - t_dur - t_from_merger, t_max - t_from_merger,
num_pts)
x_disp = np.linspace(0, max(x), num_pts)
#x = np.linspace(0, 0.3, 10000)
y = np.array([
detector_response(t, r, mc, t_coal, phi_coal, iota, theta, phi, psi)
for t in x
])
#y = np.array([f_gw(t, t_coal, mc) for t in x])
fig = plt.figure(figsize=(1980.0 / 96, 1080.0 / 96))
plt.rcParams.update({'font.size': 32, 'text.usetex': 'true'})
ax = fig.add_subplot(111)
ax.plot(x_disp, y, color='black')
ax.set_ylabel('Strain', rotation=0)
ax.set_xlabel('t')
ax.set_xticklabels([])
ax.set_yticklabels([])
ax.set_xticks([])
ax.set_yticks([])
ax.spines['left'].set_position('zero')
ax.spines['bottom'].set_position('zero')
ax.spines['right'].set_color('none')
ax.spines['top'].set_color('none')
ax.xaxis.set_label_coords(1.0, 0.44)
ax.yaxis.set_label_coords(0.0, 0.9)
#ax.set_xlim(0, 1.25 * max(x_disp))
plt.savefig(os.path.join(images_path(), 'linear_waveform.png'))
plt.show()
return
def main():
data_path = '/home/marlin/Documents/Documents/Studium/Masterarbeit/Forschungsphase/Code/Git/master_project/bns_net/saves/inception_res_net_266201913382/inception_res_net_snr_plot_last_epoch_266201913382.hf5'
with h5py.File(data_path, 'r') as FILE:
x = FILE['x1'][:]
y = FILE['y'][:]
line = np.linspace(0, 1.1 * max(x))
fig = plt.figure(figsize=(1980.0 / 96, 1080.0 / 96))
plt.rcParams.update({'font.size': 32, 'text.usetex': 'true'})
plt.grid()
plt.scatter(x, y)
plt.plot(line, line, color='red', zorder=4)
plt.xlabel('Label SNR')
plt.ylabel('Recovered SNR')
plt.savefig(os.path.join(images_path(), 'streak_plot.png'))
plt.show()
return
def main():
approximant = 'TaylorF2'
mass1 = 1.4
mass2 = 1.4
delta_t = 1.0 / 4096
f_lower = 20
final_x_sec = 68
hp, hc = get_td_waveform(approximant=approximant,
mass1=mass1,
mass2=mass2,
delta_t=delta_t,
f_lower=f_lower)
dur = hp.sample_times[-1]
hp = hp.time_slice(dur - final_x_sec - hp.delta_t, dur)
x = hp.sample_times
y = hp.data[:]
box_height = 1.1 * max(abs(y))
print("Duration: {}".format(hp.duration))
print("Length y: {}".format(len(y)))
fig = plt.figure(figsize=(1980.0 / 96, 1080.0 / 96))
plt.rcParams.update({'font.size': 32, 'text.usetex': 'true'})
ax = fig.add_subplot(111)
lw = 2
separate = 0.25
window_size = 0.5
left_edge = dur - window_size
rect_1s_1 = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='black',
facecolor='none',
zorder=4)
ax.add_patch(rect_1s_1)
window_size = 0.5
left_edge -= window_size
rect_1s_2 = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='purple',
facecolor='none',
zorder=4)
ax.add_patch(rect_1s_2)
window_size = 1.0
left_edge -= window_size
rect_2s = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='blue',
facecolor='none',
zorder=4)
ax.add_patch(rect_2s)
window_size = 2.0
left_edge -= window_size
rect_4s = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='cyan',
facecolor='none',
zorder=4)
ax.add_patch(rect_4s)
window_size = 4.0
left_edge -= window_size
rect_8s = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='green',
facecolor='none',
zorder=4)
ax.add_patch(rect_8s)
window_size = 8.0
left_edge -= window_size
rect_16s = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='yellow',
facecolor='none',
zorder=4)
ax.add_patch(rect_16s)
window_size = 16.0
left_edge -= window_size
rect_32s = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='magenta',
facecolor='none',
zorder=4)
ax.add_patch(rect_32s)
window_size = 32.0
left_edge -= window_size
rect_64s = patches.Rectangle((left_edge + separate / 2, -box_height),
window_size - separate / 2,
2 * box_height,
linewidth=lw,
edgecolor='red',
facecolor='none',
zorder=4)
ax.add_patch(rect_64s)
ax.plot(x, y)
#ax.set_ylabel('Strain', rotation=0)
ax.set_xlabel('t in s')
ax.set_yticklabels([])
ax.set_yticks([])
ax.xaxis.set_label_coords(1.05, 0)
#ax.yaxis.set_label_coords(0.05, 0.9)
ax.spines['left'].set_color('none')
ax.spines['right'].set_color('none')
ax.spines['top'].set_color('none')
ax.set_ylim((-1.2 * max(abs(y)), 1.2 * max(abs(y))))
plt.savefig(os.path.join(images_path(), 'multirate_filtering.png'))
plt.show()
return
def main():
x = np.linspace(-2, 2, 1000)
y_true = f(x)
y_para = g(x)
y_over = h(x, 5)
x_sample = np.array([-1, 0, 1])
y_sample = f(x_sample)
#Sets how many parts the figure has
gridsize = (2, 3)
#Sets the aspect-ratio of all the small plots
#0.25 is the part that sets it to have a usual aspect-ratio of 1
#The factor behind is y / x, (i.e. 3 / 4 for 4:3 final aspect-ratio)
global_aspect = 0.25 * 10.0 / 16.0
fig = plt.figure(figsize=(1980.0 / 96, 1080.0 / 96))
plt.rcParams.update({'font.size': 22, 'text.usetex': 'true'})
ax1 = plt.subplot2grid(gridsize, (0, 0), rowspan=2)
ax2 = plt.subplot2grid(gridsize, (0, 1))
ax3 = plt.subplot2grid(gridsize, (1, 1))
ax4 = plt.subplot2grid(gridsize, (0, 2))
ax5 = plt.subplot2grid(gridsize, (1, 2))
#Ground trues
ax1.set_aspect(global_aspect)
ax1.spines['left'].set_position('zero')
ax1.spines['bottom'].set_position('zero')
ax1.spines['right'].set_color('none')
ax1.spines['top'].set_color('none')
ax1.set_yticks([5, 10, 15])
ax1.plot(x, y_true, '--', color='black', zorder=1)
ax1.scatter(x_sample, y_sample, color='red', zorder=2)
#Plot arrow pointing up
ax2.set_xlim(-1, 1)
ax2.set_ylim(-1, 1)
ax2.axis('off')
x_start = -0.6
y_start = -0.22 - 0.3
dx = 2 * 0.6
dy = 2 * 0.22
ax2.arrow(x_start,
y_start,
dx,
dy,
color='black',
width=0.001,
head_width=0.075,
length_includes_head=True,
head_length=0.15)
mid_x = x_start + dx / 2
mid_y = y_start + dy / 2
ax2.text(mid_x,
mid_y,
"Quadratic fit",
ha='center',
va='top',
rotation=rad_to_deg(np.arctan(dy / dx)))
#Plot arrow pointing down
ax3.set_xlim(-1, 1)
ax3.set_ylim(-1, 1)
ax3.axis('off')
x_start = -0.6
y_start = 0.22 + 0.3
dx = 2 * 0.6
dy = -2 * 0.22
ax3.arrow(x_start,
y_start,
dx,
dy,
color='black',
width=0.001,
head_width=0.075,
length_includes_head=True,
head_length=0.15)
mid_x = x_start + dx / 2
mid_y = y_start + dy / 2
ax3.text(mid_x,
mid_y,
"Cubic fit",
ha='center',
va='bottom',
rotation=rad_to_deg(np.arctan(dy / dx)))
#Parabola regression
ax4.set_aspect(global_aspect)
ax4.spines['left'].set_position('zero')
ax4.spines['bottom'].set_position('zero')
ax4.spines['right'].set_color('none')
ax4.spines['top'].set_color('none')
ax4.set_yticks([5, 10, 15])
ax4.plot(x, y_para, zorder=1)
ax4.scatter(x_sample, y_sample, color='red', zorder=2)
#Cubic regression
ax5.set_ylim(ax4.get_ylim())
ax5.set_aspect(global_aspect)
ax5.spines['left'].set_position('zero')
ax5.spines['bottom'].set_position('zero')
ax5.spines['right'].set_color('none')
ax5.spines['top'].set_color('none')
ax5.set_yticks([5, 10, 15])
ax5.plot(x, y_true, '--', color='black', zorder=1)
ax5.plot(x, y_over, zorder=2)
ax5.scatter(x_sample, y_sample, color='red', zorder=3)
plt.savefig(os.path.join(images_path(), 'overfitting.png'))
plt.show()
return
def main():
FontSize = 48
results_path = '/home/marlin/Documents/Documents/Studium/Masterarbeit/Forschungsphase/Code/Git/master_project/bns_net/saves/long_data_2/results'
snrSensPath = os.path.join(results_path,
'new_SNR_200_steps_sensitivty.hf5')
pSensPath = os.path.join(results_path,
'p_score_200_steps_new_log_sensitivty.hf5')
snrSensSnrPath = os.path.join(results_path,
'new_SNR_200_steps_sensitivty_snr.hf5')
pSensSnrPath = os.path.join(
results_path, 'p_score_200_steps_new_log_sensitivty_snr.hf5')
#combinedSnrFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#combinedPFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#Plot sensitivity as range
dpi = 96
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
fig, ax = plt.subplots(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(snrSensPath, 'r') as f:
x = f['x'][:]
y = f['radius'][:]
with h5py.File(pSensPath, 'r') as f:
x2 = f['x'][1:-1]
y2 = f['radius'][1:-1]
ax.semilogx(x, y, label='SNR', linewidth=3)
ax.semilogx(x2, y2, label='p-score', linewidth=3)
ax.set_xlabel('False alarm per month')
ax.set_ylabel('Sensitive distance in Mpc')
x_low, x_high = ax.get_xlim()
ax.set_xlim(x_high, 9 * 10**-1)
ax.grid()
ax.legend()
plotName = 'SensitivitySNR.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot sensitivity as percentage
dpi = 96
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
fig, ax = plt.subplots(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(snrSensSnrPath, 'r') as f:
x = f['bins'][:]
y = f['y'][:]
maxIdx = 56 - min(x) + 1
x = x[:maxIdx] / (39.225 * np.sqrt(2))
y = y[:maxIdx]
with h5py.File(pSensSnrPath, 'r') as f:
x2 = f['bins'][:]
y2 = f['y'][:]
x2 = x2[:maxIdx] / (39.225 * np.sqrt(2))
y2 = y2[:maxIdx]
ax.plot(x, y, label='SNR', linewidth=3)
ax.plot(x2, y2, label='p-score', linewidth=3)
ax.set_xlabel('pSNR')
ax.set_ylabel('Ratio of detected signals')
#x_low, x_high = ax.get_xlim()
#ax.set_xlim(x_high, 10**-1)
ax.grid()
ax.legend()
plotName = 'SensitivityPercentage.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
return
def main():
x = np.linspace(-4, 4, 10000)
y_1 = np.array([Err_1(pt) for pt in x])
y_2 = np.array([Err_2(pt) for pt in x])
y_3 = np.array([Err_3(pt) for pt in x])
gridsize = (6, 3)
#Sets the aspect-ratio of all the small plots
#0.25 is the part that sets it to have a usual aspect-ratio of 1
#The factor behind is y / x, (i.e. 3 / 4 for 4:3 final aspect-ratio)
global_aspect = 0.25 * 10.0 / 16.0
fig = plt.figure(figsize=(1980.0/96, 1080.0/96))
plt.rcParams.update({'font.size': 22, 'text.usetex': 'true'})
ax1 = plt.subplot2grid(gridsize, (0, 0), rowspan=5)
ax2 = plt.subplot2grid(gridsize, (0, 1), rowspan=5)
ax3 = plt.subplot2grid(gridsize, (0, 2), rowspan=5)
ax4 = plt.subplot2grid(gridsize, (5, 0))
ax5 = plt.subplot2grid(gridsize, (5, 1))
ax6 = plt.subplot2grid(gridsize, (5, 2))
#Plot non squished error
ax1.plot(x, y_1)
ax1.grid()
ax1.set_xlabel('x')
ax1.set_ylabel('y')
ax1.set_xticks([-4, -2, 0, 2, 4])
#Plot squished error
ax2.plot(x, y_2)
ax2.grid()
ax2.set_xlabel('x')
ax2.set_ylabel('y')
ax2.set_xticks([-4, -2, 0, 2, 4])
#Plot squished and cutoff error
ax3.plot(x, y_3)
ax3.grid()
ax3.set_xlabel('x')
ax3.set_ylabel('y')
ax3.set_xticks([-4, -2, 0, 2, 4])
#Put label (a)
ax4.set_xlim(-1, 1)
ax4.set_ylim(-1, 1)
ax4.axis('off')
ax4.text(0, 0, '(a)', ha='center', va='bottom')
#Put label (b)
ax5.set_xlim(-1, 1)
ax5.set_ylim(-1, 1)
ax5.axis('off')
ax5.text(0, 0, '(b)', ha='center', va='bottom')
#Put label (c)
ax6.set_xlim(-1, 1)
ax6.set_ylim(-1, 1)
ax6.axis('off')
ax6.text(0, 0, '(c)', ha='center', va='bottom')
plt.tight_layout()
plt.savefig(os.path.join(images_path(), 'loss_evolution.png'))
plt.show()
return
def main():
FontSize = 48
results_path = '/home/marlin/Documents/Documents/Studium/Masterarbeit/Forschungsphase/Code/Git/master_project/bns_net/saves/long_data_2/results'
snrSensPath = os.path.join(
results_path, 'combined_steps_P_log_scale_SNR_sensitivity.hf5')
pSensPath = os.path.join(results_path,
'combined_steps_P_log_scale_sensitivity.hf5')
#combinedSnrFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#combinedPFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#Plot SNR sensitivity as range
with h5py.File(snrSensPath, 'r') as f:
x = f['snr'][:]
y = f['p-score'][:]
z = f['radius'][:]
fa = f['xy'][:]
per = f['percentage'][:]
shape = (int(np.sqrt(len(x))), int(np.sqrt(len(x))))
x = x.reshape(shape)
y = y.reshape(shape)
z = z.reshape(shape)
fa = fa.reshape(shape)
per = per.reshape(shape)
snr = x[0]
p = y.transpose()[0]
dpi = 96
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
fig, ax = plt.subplots(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
ps = [0.408, 0.5, 0.75, 0.95]
colors = ['blue', 'red', 'green', 'black']
for i, pt in enumerate(ps):
minIdx = np.argmin(np.abs(p - pt))
lab = p[minIdx]
ax.semilogx(fa[minIdx][:-1],
z[minIdx][:-1],
label='p-score: %.2f' % lab,
color=colors[i])
ax.set_xlabel('False alarms per month')
ax.set_ylabel('Sensitive distance in Mpc')
x_low, x_high = ax.get_xlim()
ax.set_xlim(x_high, 9 * 10**-1)
#plt.plot([min(snr), max(snr)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
ax.grid()
ax.legend()
plotName = 'CombinedSensitivitySNR.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot SNR percentages
dpi = 96
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
fig, ax = plt.subplots(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
for i, pt in enumerate(ps):
minIdx = np.argmin(np.abs(p - pt))
lab = p[minIdx]
ax.semilogx(fa[minIdx],
per[minIdx],
label='p-score: %.2f' % lab,
color=colors[i])
ax.set_xlabel('False alarms per month')
ax.set_ylabel('Ratio of detected signals')
x_low, x_high = ax.get_xlim()
ax.set_xlim(x_high, 10**-1)
#plt.plot([min(snr), max(snr)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
ax.grid()
ax.legend()
plotName = 'CombinedSensitivityPercentageSNR.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot p-score sensitivity as range
with h5py.File(snrSensPath, 'r') as f:
x = f['snr'][:]
y = f['p-score'][:]
z = f['radius'][:]
fa = f['xy'][:]
per = f['percentage'][:]
shape = (int(np.sqrt(len(x))), int(np.sqrt(len(x))))
x = x.reshape(shape)
y = y.reshape(shape)
z = z.reshape(shape)
fa = fa.reshape(shape)
per = per.reshape(shape)
snr = x[0]
p = y.transpose()[0]
dpi = 96
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
fig, ax = plt.subplots(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
snrs = [4, 6.7, 8, 15]
colors = ['blue', 'red', 'green', 'black']
for i, pt in enumerate(snrs):
minIdx = np.argmin(np.abs(snr - pt))
lab = snr[minIdx]
ax.semilogx(fa.transpose()[minIdx][:-1],
z.transpose()[minIdx][:-1],
label='SNR: %.2f' % lab,
color=colors[i])
ax.set_xlabel('False alarms per month')
ax.set_ylabel('Sensitive distance in Mpc')
x_low, x_high = ax.get_xlim()
ax.set_xlim(x_high, 9 * 10**-1)
#plt.plot([min(snr), max(snr)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
ax.grid()
ax.legend()
plotName = 'CombinedSensitivityP.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot p-score percentages
dpi = 96
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
fig, ax = plt.subplots(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
for i, pt in enumerate(snrs):
minIdx = np.argmin(np.abs(snr - pt))
lab = snr[minIdx]
ax.semilogx(fa.transpose()[minIdx],
per.transpose()[minIdx],
label='SNR: %.2f' % lab,
color=colors[i])
ax.set_xlabel('False alarms per month')
ax.set_ylabel('Ratio of detected signals')
x_low, x_high = ax.get_xlim()
ax.set_xlim(x_high, 10**-1)
#plt.plot([min(snr), max(snr)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
ax.grid()
ax.legend()
plotName = 'CombinedSensitivityPercentageP.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
return
def main():
FontSize = 48
results_path = '/home/marlin/Documents/Documents/Studium/Masterarbeit/Forschungsphase/Code/Git/master_project/bns_net/saves/long_data_2/results'
snrFalseAlarmPath = os.path.join(results_path,
'SNR_200_steps_false_alarm.hf5')
pFalseAlarmPath = os.path.join(
results_path, 'p_score_200_steps_new_log_false_alarm.hf5')
#combinedSnrFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#combinedPFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#Plot SNR false alarm rate
dpi = 96
plt.figure(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(snrFalseAlarmPath, 'r') as f:
x = f['x'][:]
y = f['y'][:]
plt.semilogy(x, y, label='Our network')
plt.xlabel('SNR')
plt.ylabel('False alarms per month')
plt.plot([min(x), max(x)], [15500, 15500],
color='black',
linestyle='dotted',
label='Upper limit Krastev',
linewidth=2)
y_low, y_high = plt.ylim()
plt.ylim(9 * 10**-1, y_high)
plt.grid()
plt.legend()
plotName = 'FalseAlarmSNR.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot p-score false alarm rate
dpi = 96
plt.figure(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(pFalseAlarmPath, 'r') as f:
x = f['x'][1:]
y = f['y'][1:]
plt.semilogy(x, y, label='Our network')
plt.xlabel('p-score')
plt.ylabel('False alarms per month')
plt.plot([min(x), max(x)], [15500, 15500],
color='black',
linestyle='dotted',
label='Upper limit Krastev',
linewidth=2)
y_low, y_high = plt.ylim()
plt.ylim(9 * 10**-1, y_high)
plt.grid()
plt.legend()
plotName = 'FalseAlarmP.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot p-score false alarm rate 2
dpi = 96
plt.figure(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(pFalseAlarmPath, 'r') as f:
x = -np.log(1 - f['x'][1:])
y = f['y'][1:]
plt.loglog(x, y, label='Our network')
plt.plot([plt.xlim()[0], plt.xlim()[1]], [15500, 15500],
color='black',
linestyle='dotted',
label='Upper limit Krastev',
linewidth=2)
plt.xlabel('$-\log (1-$p-score$)$')
plt.ylabel('False alarms per month')
y_low, y_high = plt.ylim()
plt.ylim(9 * 10**-1, y_high)
plt.grid()
#plt.ylim(plt.ylim()[0], 100000)
plt.legend()
plotName = 'FalseAlarmP2.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
return
def main():
FontSize = 48
results_path = '/home/marlin/Documents/Documents/Studium/Masterarbeit/Forschungsphase/Code/Git/master_project/bns_net/saves/long_data_2/results'
snrFalseAlarmPath = os.path.join(
results_path, 'combined_steps_P_log_scale_SNR_false_alarm.hf5')
pFalseAlarmPath = os.path.join(
results_path, 'combined_steps_P_log_scale_false_alarm.hf5')
#combinedSnrFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#combinedPFalseAlarmPath = os.path.join(results_path, 'combined_snr_100_steps.hf5')
#Plot SNR false alarm rate
dpi = 96
plt.figure(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(snrFalseAlarmPath, 'r') as f:
x = f['x'][:]
y = f['y'][:]
z = f['z'][:]
shape = (int(np.sqrt(len(x))), int(np.sqrt(len(x))))
x = x.reshape(shape)
y = y.reshape(shape)
z = z.reshape(shape)
snr = x[0]
p = y.transpose()[0]
ps = [0.408, 0.5, 0.75, 0.95]
colors = ['blue', 'red', 'green', 'black']
for i, pt in enumerate(ps):
minIdx = np.argmin(np.abs(p - pt))
lab = p[minIdx]
plt.semilogy(x[0],
z[minIdx],
label='p-score: %.2f' % lab,
color=colors[i])
print("False alarm rate SNR at p-score: %.2f" % lab)
print("Max (x, y): ({}, {})".format(x[0][0], z[minIdx][0]))
print("Min (x, y): ({}, {})".format(x[0][-1], z[minIdx][-1]))
plt.xlabel('SNR')
plt.ylabel('False alarms per month')
#plt.plot([min(snr), max(snr)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
y_low, y_high = plt.ylim()
#plt.ylim(10**-1, y_high)
plt.ylim(9 * 10**-1, 2 * 10**3)
plt.grid()
plt.legend()
plotName = 'CombinedFalseAlarmSNR.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot p-score false alarm rate
dpi = 96
plt.figure(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(pFalseAlarmPath, 'r') as f:
x = f['x'][:]
y = f['y'][:]
z = f['z'][:]
shape = (int(np.sqrt(len(x))), int(np.sqrt(len(x))))
x = x.reshape(shape)
y = y.reshape(shape)
z = z.reshape(shape)
snr = x[0]
p = y.transpose()[0]
snrs = [4, 6.7, 8, 15]
colors = ['blue', 'red', 'green', 'black']
for i, pt in enumerate(snrs):
minIdx = np.argmin(np.abs(snr - pt))
lab = snr[minIdx]
plt.semilogy(y.transpose()[0],
z.transpose()[minIdx],
label='SNR: %.2f' % lab,
color=colors[i])
print("False alarm rate p-score at SNR: %.2f" % lab)
print("Max (x, y): ({}, {})".format(y.transpose()[0][0],
z.transpose()[minIdx][0]))
print("Min (x, y): ({}, {})".format(y.transpose()[0][-2],
z.transpose()[minIdx][-2]))
plt.xlabel('p-score')
plt.ylabel('False alarms per month')
#plt.plot([min(p), max(p)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
y_low, y_high = plt.ylim()
#plt.ylim(10**-1, y_high)
plt.ylim(9 * 10**-1, 2 * 10**3)
plt.grid()
plt.legend()
plotName = 'CombinedFalseAlarmP.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
#Plot p-score false alarm rate 2
dpi = 96
plt.figure(figsize=(1920.0 / dpi, 1440.0 / dpi), dpi=dpi)
plt.rcParams.update({'font.size': FontSize, 'text.usetex': 'true'})
with h5py.File(pFalseAlarmPath, 'r') as f:
x = f['x'][:]
y = -np.log(1 - f['y'][:])
z = f['z'][:]
shape = (int(np.sqrt(len(x))), int(np.sqrt(len(x))))
x = x.reshape(shape)
y = y.reshape(shape)
z = z.reshape(shape)
snr = x[0]
p = y.transpose()[0]
snrs = [4, 6.7, 8, 15]
colors = ['blue', 'red', 'green', 'black']
for i, pt in enumerate(snrs):
minIdx = np.argmin(np.abs(snr - pt))
lab = snr[minIdx]
plt.loglog(y.transpose()[0],
z.transpose()[minIdx],
label='SNR: %.2f' % lab,
color=colors[i])
plt.xlabel('$-\log (1-$p-score$)$')
plt.ylabel('False alarms per month')
#plt.plot([min(p), max(p)], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
y_low, y_high = plt.ylim()
#plt.ylim(9*10**-1, y_high)
plt.ylim(9 * 10**-1, 2 * 10**3)
plt.grid()
plt.legend()
plotName = 'CombinedFalseAlarmP2.png'
plt.savefig(os.path.join(images_path(), plotName))
plt.show()
##Plot p-score false alarm rate 2
#dpi = 96
#plt.figure(figsize=(1920.0/dpi, 1440.0/dpi), dpi=dpi)
#plt.rcParams.update({'font.size': 32, 'text.usetex': 'true'})
#with h5py.File(pFalseAlarmPath, 'r') as f:
#x = -np.log(1-f['x'][1:])
#y = f['y'][1:]
#plt.loglog(x, y, label='Our network')
#plt.plot([plt.xlim()[0], plt.xlim()[1]], [15500, 15500], color='black', linestyle='dotted', label='Upper limit Krastev', linewidth=2)
#plt.xlabel('$-\log (1-$p-score$)$')
#plt.ylabel('False alarms per month')
#plt.grid()
##plt.ylim(plt.ylim()[0], 100000)
#plt.legend()
#plotName = 'CombinedFalseAlarmP2.png'
#plt.savefig(os.path.join(images_path(), plotName))
#plt.show()
return