fig, ax = plt.subplots(1, 1)
for i, target in enumerate(targets):
print('doing channel %s' % target)
inChannels_single = np.array(['I1V', 'I2V', 'I3V', 'MWI-15', 'MWI-16'])
file_single = 'qrnn_ici_%s_%s_%s_mwi-alone.nc' % (depth, width, target)
print(file_single)
i183, = np.argwhere(inChannels_single == target)[0]
print(i183)
y_pre, y_prior, y0, y, y_pos_mean =\
read_qrnn(file_single, inChannels_single, target)
im = np.abs(y_pre[:, 3] - y_prior[:, i183]) <= 7.5
# im = np.abs(y_prior[:, i183]- y0) <= 7.5
print('rejected QRNN ', (1 - np.sum(im) / im.size) * 100)
bia = stats.calculate_bias(y_prior, y0, y, y_pre[:, 3], im, i183)
std = stats.calculate_std(y_prior, y0, y, y_pre[:, 3], im, i183)
ske = stats.calculate_skew(y_prior, y0, y, y_pre[:, 3], im, i183)
mae = stats.calculate_mae(y_prior, y0, y, y_pre[:, 3], im, i183)
print('rejected B183 ', (np.sum(im_b) / im_b.size) * 100)
bia_b = stats.calculate_bias(y_prior, y0, y, y_pre[:, 3], ~im_b, i183)
std_b = stats.calculate_std(y_prior, y0, y, y_pre[:, 3], ~im_b, i183)
ske_b = stats.calculate_skew(y_prior, y0, y, y_pre[:, 3], ~im_b, i183)
mae_b = stats.calculate_mae(y_prior, y0, y, y_pre[:, 3], ~im_b, i183)
#%%
bia = list(bia + bia_b)
mae = list(mae + mae_b)
ske = list(ske + ske_b)
std = list(std + std_b)
#%%
for i, target in enumerate(targets):
inChannels_single = np.array([target, 'I5V' , 'I6V', 'I7V', 'I8V', \
'I9V', 'I10V', 'I11V'])
file_single = 'qrnn_ici_%s_%s_%s_single.nc' % (depth, width, target)
print(file_single)
i183, = np.argwhere(inChannels_single == target)[0]
y_pre, y_prior, y0, y, y_pos_mean = read_qrnn(file_single,\
inChannels_single, target )
im = np.abs(y_pre[:, 3] - y_prior[:, i183]) >= 10
print((np.sum(im) / im.size) * 100)
bia = stats.calculate_bias(y_prior, y0, y, y_pre[:, 3], im, i183)
std = stats.calculate_std(y_prior, y0, y, y_pre[:, 3], im, i183)
ske = stats.calculate_skew(y_prior, y0, y, y_pre[:, 3], im, i183)
mae = stats.calculate_mae(y_prior, y0, y, y_pre[:, 3], im, i183)
inChannels_all = np.array(['I1V', 'I2V', 'I3V', 'I5V' , 'I6V', 'I7V', \
'I8V', 'I9V', 'I10V', 'I11V'])
file_all = 'qrnn_ici_%s_%s_%s.nc' % (depth, width, target)
print(file_all)
i183, = np.argwhere(inChannels_all == target)[0]
y_pre, y_prior1, y01, y, y_pos_mean1 = read_qrnn(file_all,\
inChannels_all, target )
im = np.abs(y_pre[:, 3] - y_prior[:, i183]) >= 10
print((np.sum(im) / im.size) * 100)
ax2 = ax.twinx()
for i, target in enumerate(targets):
inChannels_single = np.array(
[target, 'I5V', 'I6V', 'I7V', 'I8V', 'I9V', 'I10V', 'I11V'])
file_single = 'qrnn_ici_%s_%s_%s_single.nc' % (depth, width, target)
print(file_single)
i183, = np.argwhere(inChannels_single == target)[0]
y_pre, y_prior, y0, y, y_pos_mean = read_qrnn(file_single,
inChannels_single, target)
bias = ()
rej = []
for j in filters:
im = np.abs(y_pre[:, 3] - y_prior[:, i183]) <= j
bias += stats.calculate_bias(y_prior, y0, y, y_pre[:, 3], im, i183)
rej.append((1 - np.sum(im) / im.size) * 100)
bias = np.array(list(bias))
# rej = list(rej)
color = 'tab:red'
ax.plot( bias[[4, 9, 14, 19, 24, 29, 34, 39, 44, 49]], color = color,\
linewidth = 2.5, linestyle = linetype[i])
ax.set_ylim(-0.35, 0)
# instantiate a second axes that shares the same x-axis
color = 'tab:blue'
ax2.plot(rej, linestyle = linetype[i], color = color, \
linewidth = 2.5)