def PlotPossibleGeoSpace(figname, Xpred_dir, compare_original=False):
"""
Function to plot the possible geometry space for a model evaluation result.
It reads from Xpred_dir folder and finds the Xpred result insdie and plot that result
:params figname: The name of the figure to save
:params Xpred_dir: The directory to look for Xpred file which is the source of plotting
:output A plot containing 4 subplots showing the 8 geomoetry dimensions
"""
Xpredfile = os.path.join(Xpred_dir,
get_pred_truth_file.get_Xpred(Xpred_dir))
Xpred = pd.read_csv(Xpredfile, header=None, delimiter=' ').values
Xtruthfile = os.path.join(Xpred_dir,
get_pred_truth_file.get_Xtruth(Xpred_dir))
Xtruth = pd.read_csv(Xtruthfile, header=None, delimiter=' ').values
f = plt.figure()
ax0 = plt.gca()
print(np.shape(Xpred))
#print(Xpred)
plt.title(figname)
for i in range(4):
ax = plt.subplot(2, 2, i + 1)
ax.scatter(Xpred[:, i], Xpred[:, i + 4], label="Xpred")
if (compare_original):
ax.scatter(Xtruth[:, i], Xtruth[:, i + 4], label="Xtruth")
plt.xlabel('h{}'.format(i))
plt.ylabel('r{}'.format(i))
plt.xlim(-1, 1)
plt.ylim(-1, 1)
plt.legend()
#plt.title(figname)
f.savefig(figname + '.png')
def PlotPairwiseGeometry(figname, Xpred_dir):
"""
Function to plot the pair-wise scattering plot of the geometery file to show
the correlation between the geometry that the network learns
"""
Xpredfile = get_pred_truth_file.get_Xpred(Xpred_dir)
Xpred = pd.read_csv(Xpredfile, header=None, delimiter=' ')
f = plt.figure()
axes = pd.plotting.scatter_matrix(Xpred, alpha=0.2)
#plt.tight_layout()
plt.title("Pair-wise scattering of Geometery predictions")
plt.savefig(figname)
def after_Tandem_pred():
"""
This function handles the rest of the evaluation after the Ypred has been generated by the Tandem model (forward model)
"""
data_dir = 'data'
Ypred_file = get_pred_truth_file.get_Ypred(data_dir)
Xpred_file = get_pred_truth_file.get_Xpred(data_dir)
Ytruth_file = get_pred_truth_file.get_Ytruth(data_dir)
Ypred = np.loadtxt(Ypred_file, delimiter=' ')
Ytruth = np.loadtxt(Ytruth_file, delimiter=' ')
Xpred = np.loadtxt(Xpred_file, delimiter=' ')
l_Ypred = len(Ypred)
l_Ytruth = len(Ytruth)
k = l_Ypred / l_Ytruth
print("l_Ypred", l_Ypred)
print("l_Ytruth", l_Ytruth)
print("k", k)
assert k - int(
k) < 0.001, "Check you length, the divide result k is not an int!!"
print(
'For each data point in your truth file, the VAE generated {} data points'
.format(k))
k = int(k)
#best_index_list = np.zeros([1,l_Ytruth])
Xpred_new = np.zeros([l_Ytruth, 8])
Ypred_new = np.zeros(np.shape(Ytruth))
for i in range(l_Ytruth):
diff_mat = Ypred[i * k:(i + 1) * k, :] - Ytruth[i, :]
distance_mat = np.linalg.norm(diff_mat, axis=1)
best_index = np.argmax(distance_mat)
#best_index_list[i] = best_index
Xpred_new[i, :] = Xpred[i * k + best_index, :]
Ypred_new[i, :] = Ypred[i * k + best_index, :]
with open(Xpred_file, 'w') as f1:
np.savetxt(f1, Xpred_new, fmt='%.3f')
with open(Ypred_file, 'w') as f2:
np.savetxt(f2, Ypred_new, fmt='%.3f')
mae, mse = compare_truth_pred(Ypred_file, Ytruth_file)
plt.figure(figsize=(12, 6))
plt.hist(mse, bins=100)
plt.xlabel('Mean Squared Error')
plt.ylabel('cnt')
plt.suptitle('VAE (Avg MSE={:.4e})'.format(np.mean(mse)))
plt.savefig(os.path.join(os.path.abspath(''), 'data', 'VAE.png'))
plt.show()
print('VAE (Avg MSE={:.4e})'.format(np.mean(mse)))
def PlotPossibleGeoSpace(figname,
Xpred_dir,
compare_original=False,
calculate_diversity=None):
"""
Function to plot the possible geometry space for a model evaluation result.
It reads from Xpred_dir folder and finds the Xpred result insdie and plot that result
:params figname: The name of the figure to save
:params Xpred_dir: The directory to look for Xpred file which is the source of plotting
:output A plot containing 4 subplots showing the 8 geomoetry dimensions
"""
Xpredfile = get_pred_truth_file.get_Xpred(Xpred_dir)
Xpred = pd.read_csv(Xpredfile, header=None, delimiter=' ').values
Xtruthfile = get_pred_truth_file.get_Xtruth(Xpred_dir)
Xtruth = pd.read_csv(Xtruthfile, header=None, delimiter=' ').values
f = plt.figure()
ax0 = plt.gca()
print(np.shape(Xpred))
#print(Xpred)
#plt.title(figname)
if (calculate_diversity == 'MST'):
diversity_Xpred, diversity_Xtruth = calculate_MST(Xpred, Xtruth)
elif (calculate_diversity == 'AREA'):
diversity_Xpred, diversity_Xtruth = calculate_AREA(Xpred, Xtruth)
for i in range(4):
ax = plt.subplot(2, 2, i + 1)
ax.scatter(Xpred[:, i], Xpred[:, i + 4], s=3, label="Xpred")
if (compare_original):
ax.scatter(Xtruth[:, i], Xtruth[:, i + 4], s=3, label="Xtruth")
plt.xlabel('h{}'.format(i))
plt.ylabel('r{}'.format(i))
plt.xlim(-1, 1)
plt.ylim(-1, 1)
plt.legend()
if (calculate_diversity != None):
plt.text(-4,
3.5,
'Div_Xpred = {}, Div_Xtruth = {}, under criteria {}'.format(
diversity_Xpred, diversity_Xtruth, calculate_diversity),
zorder=1)
plt.suptitle(figname)
f.savefig(figname + '.png')