def visualize_2d(save_path, embedding, label, domain, class_num):
fig = plt.figure()
ax = fig.add_subplot(111)
colors = cm.rainbow(np.linspace(0.0, 1.0, class_num))
xx = embedding[:, 0]
yy = embedding[:, 1]
for i in range(class_num):
ax.scatter(xx[label == i], yy[label == i], color=colors[i], s=10)
'''
for i in range(embedding.shape[0]):
plt.text(xx[i], yy[i], str(label[i]), fontdict={"size": 10})
'''
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig(os.path.join(save_path, "TSNE_Label_2D.pdf"),
format='pdf',
dpi=600)
plt.show()
plt.close()
fig = plt.figure()
ax = fig.add_subplot(111)
colors = cm.rainbow(np.linspace(0.0, 1.0, class_num))
for i in range(2):
ax.scatter(xx[domain == i],
yy[domain == i],
color=cm.bwr(i / 1.),
s=10)
'''
for i in range(embedding.shape[0]):
plt.text(xx[i], yy[i], str(int(domain[i])), fontdict={
"size": 10}, color=cm.bwr(domain[i]/1.))
'''
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig(os.path.join(save_path, "TSNE_Domain_2D.pdf"),
format='pdf',
dpi=600)
plt.show()
plt.close()
def visualize_3d(save_path, embedding, label, domain, class_num):
fig = plt.figure()
ax = fig.add_subplot(111, projection="3d")
colors = cm.rainbow(np.linspace(0.0, 1.0, class_num))
xx = embedding[:, 0]
yy = embedding[:, 1]
zz = embedding[:, 2]
for i in range(class_num):
ax.scatter(xx[label == i],
yy[label == i],
zz[label == i],
color=colors[i],
s=10)
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
ax.zaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig(os.path.join(save_path, "TSNE_Label_3D.pdf"),
format='pdf',
dpi=600)
plt.show()
plt.close()
fig = plt.figure()
ax = fig.add_subplot(111, projection="3d")
for i in range(2):
ax.scatter(xx[domain == i],
yy[domain == i],
zz[domain == i],
color=cm.bwr(i / 1.),
s=10)
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
ax.zaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig(os.path.join(save_path, "TSNE_Domain_3D.pdf"),
format='pdf',
dpi=600)
plt.show()
plt.close()
def graph_combi(self,par,dico,lar,lon):
""" This method produces a plot as a cumulative plot where the results
of the Maximum Likehood Method are directly shown in order to allow the
user to checked graphically (GSA) if the number of populations is adapted.
If not he can adapt the number of populations """
fig2 = plt.figure(2)
fig2.canvas.manager.window.resizable(int(lar/2), int(lon/2))
fig2.canvas.manager.window.wm_geometry("+"+str(int(lon/2))+"+0")
ax = fig2.add_subplot(111)
lis=list(self.df[self.parameter].dropna())
(quantiles, values), (slope, intercept, r) = stats.probplot(lis, dist='norm')
index = 0
lo = len(lis)
plt.plot(quantiles, values,'ok') # To plot the cumulative plot
color=cm.bwr(np.linspace(0,1,len(dico.keys()))) # Color for each population
for key,c in zip(dico.keys(),color):
ax.plot(quantiles[index:int(index+len(dico[key]))], values[index:int(index+len(dico[key]))],marker='o',c=c,linewidth=0) # Print the population and mixing categories
if len(dico[key])>1:
nb = len(values[index:int(index+len(dico[key]))])
perc = round(100*float(nb)/float(lo),1)
av = round(np.mean(values[index:int(index+len(dico[key])-1)]),1)
st = round(np.std(values[index:int(index+len(dico[key])-1)]),1)
else: # If there is one value for the key to avoid numpy Warning on the mean, deviation ...
nb = 1
perc = round(100*float(nb)/float(lo),1)
av = round(values[index],1)
st = 0
ax.annotate(str(key)+' ('+str(perc)+'%): Mean='+str(av)+'+/-'+str(st), xy=(quantiles[int(index+len(dico[key])/2)], values[int(index+len(dico[key])/2)]), xytext=(quantiles[int(index+len(dico[key])/2)]+0.5, values[int(index+len(dico[key])/2)]),color=c,fontsize=12) # Plot the name of the key and the mean+stdev directly on the plot
index = int(index+len(dico[key]))
ticks_perc=[1, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 95, 99] #define ticks
ticks_quan=[stats.norm.ppf(i/100.) for i in ticks_perc] #transfrom them from precentile to cumulative density
plt.xticks(ticks_quan,ticks_perc,fontsize=15) #assign new ticks
plt.xlabel('Cumulative (%)',fontsize=15)
plt.yticks(fontsize=15)
plt.ylabel('Parameter (values)',fontsize=15)
plt.title(str(par),fontsize=20)
plt.grid() #show plot
fig2.show()
def visualize(self, dim):
self.src_extractor.cpu()
self.tar_extractor.cpu()
self.src_extractor.eval()
self.tar_extractor.eval()
src_data = torch.FloatTensor()
src_label = torch.LongTensor()
for index, src in enumerate(self.src_loader):
data, label = src
src_data = torch.cat((src_data, data))
src_label = torch.cat((src_label, label))
tar_data = torch.FloatTensor()
tar_label = torch.IntTensor()
for index, tar in enumerate(self.tar_loader):
data, label = tar
tar_data = torch.cat((tar_data, data))
tar_label = torch.cat((tar_label, label))
if src_data.shape[1] != 3:
src_data = src_data.expand(
src_data.shape[0], 3, self.img_size, self.img_size)
src_data, src_label = src_data[0:1000], src_label[0:1000]
tar_data, tar_label = tar_data[0:1000], tar_label[0:1000]
src_z = self.src_extractor(src_data)
tar_z = self.src_extractor(tar_data)
data = np.concatenate(
(src_z.detach().numpy(), tar_z.detach().numpy()))
label = np.concatenate(
(src_label.numpy(), tar_label.numpy()))
tsne = TSNE(n_components=dim, verbose=1,
init="pca", perplexity=40, n_iter=3000)
embedding = tsne.fit_transform(data)
embedding_max, embedding_min = np.max(
embedding, 0), np.min(embedding, 0)
embedding = (embedding-embedding_min)/(embedding_max-embedding_min)
if dim == 2:
fig = plt.figure()
ax = fig.add_subplot(111)
colors = cm.rainbow(np.linspace(0.0, 1.0, self.num_classes))
xx = embedding[:, 0]
yy = embedding[:, 1]
for i in range(self.num_classes):
ax.scatter(xx[label == i], yy[label == i],
color=colors[i], s=10)
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig("TSNE_Label_2D.pdf", format='pdf', dpi=600)
plt.show()
plt.close("all")
src_tag = torch.zeros(src_z.size(0))
tar_tag = torch.ones(tar_z.size(0))
tag = np.concatenate((src_tag.numpy(), tar_tag.numpy()))
fig = plt.figure()
ax = fig.add_subplot(111)
colors = cm.rainbow(np.linspace(0.0, 1.0, self.num_classes))
xx = embedding[:, 0]
yy = embedding[:, 1]
for i in range(2):
ax.scatter(xx[tag == i], yy[tag == i],
color=cm.bwr(i/1.), s=10)
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig("TSNE_Domain_2D.pdf", format='pdf', dpi=600)
plt.show()
plt.close("all")
elif dim == 3:
fig = plt.figure()
ax = fig.add_subplot(111, projection="3d")
colors = cm.rainbow(np.linspace(0.0, 1.0, self.num_classes))
xx = embedding[:, 0]
yy = embedding[:, 1]
zz = embedding[:, 2]
for i in range(self.num_classes):
ax.scatter(xx[label == i], yy[label == i],
zz[label == i], color=colors[i], s=10)
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
ax.zaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig("TSNE_Label_3D.pdf", format='pdf', dpi=600)
plt.show()
plt.close("all")
src_tag = torch.zeros(src_z.size(0))
tar_tag = torch.ones(tar_z.size(0))
tag = np.concatenate((src_tag.numpy(), tar_tag.numpy()))
fig = plt.figure()
ax = fig.add_subplot(111, projection="3d")
xx = embedding[:, 0]
yy = embedding[:, 1]
zz = embedding[:, 2]
for i in range(2):
ax.scatter(xx[tag == i], yy[tag == i],
zz[tag == i], color=cm.bwr(i/1.), s=10)
ax.xaxis.set_major_formatter(NullFormatter())
ax.yaxis.set_major_formatter(NullFormatter())
ax.zaxis.set_major_formatter(NullFormatter())
plt.axis('tight')
plt.legend(loc='best', scatterpoints=1, fontsize=5)
plt.savefig("TSNE_Domain_3D.pdf", format='pdf', dpi=600)
plt.show()
plt.close("all")
def plot_campaign_pop(self, par, dictiopopu, coordonnee, lar, lon):
""" This method creates a map of the data serie architectured with
populations and mixing groups as defined above """
fig4 = plt.figure()
fig4.canvas.manager.window.resizable(int(lar / 2), int(lon / 2))
fig4.canvas.manager.window.wm_geometry("+" + str(int(lon / 2)) + "+" +
str(int(lar / 2)))
ax = fig4.add_subplot(111)
try:
dirname, filename = os.path.split(os.path.abspath(sys.argv[0]))
f = os.path.join(dirname, "Map.tif")
datafile = cbook.get_sample_data(f)
img = imread(datafile)
except:
pass
dicofinal = OrderedDict()
for key in dictiopopu.keys():
try:
minimum = np.min(dictiopopu[key])
maximum = np.max(dictiopopu[key])
dicofinal.setdefault(key, [[], []])
for i in range(0, len(self.df.values)):
if minimum <= float(
self.df[self.parameter].iloc[i]) <= maximum:
dicofinal[key][0].append(
float(self.df['LONGITUDE'].iloc[i]))
dicofinal[key][1].append(
float(self.df['LATITUDE'].iloc[i]))
else:
None
except ValueError:
pass
colors = cm.bwr(np.linspace(0, 1, len(dicofinal.keys())))
for keyf, c in zip(dicofinal.keys(), colors):
ax.scatter(dicofinal[keyf][0],
dicofinal[keyf][1],
edgecolors='black',
linewidth=1,
color=c,
marker='o',
s=50,
label=str(keyf) + ': from ' +
str(round(np.min(dictiopopu[keyf]), 2)) + ' to ' +
str(round(np.max(dictiopopu[keyf]), 2)))
handles, labels = ax.get_legend_handles_labels()
ax.legend(reversed(handles),
reversed(labels),
loc='lower left',
scatterpoints=1,
fontsize=12)
ax.ticklabel_format(useOffset=False)
plt.xticks(rotation=70)
try:
plt.imshow(img, zorder=0, extent=coordonnee)
except:
pass
plt.xlim(float(coordonnee[0]), float(coordonnee[1]))
plt.ylim(float(coordonnee[2]), float(coordonnee[3]))
plt.xlabel('Longitude', fontsize=15)
plt.ylabel('Latitude', fontsize=15)
plt.xticks(fontsize=15)
plt.yticks(fontsize=15)
props = dict(boxstyle='round', facecolor='wheat', alpha=0.5)
plt.text(0.05,
0.95,
'X1: ' + str(round(coordonnee[0], 5)) + '\n' + 'X2: ' +
str(round(coordonnee[1], 5)) + '\n' + 'Y1: ' +
str(round(coordonnee[2], 5)) + '\n' + 'Y2: ' +
str(round(coordonnee[3], 5)),
transform=ax.transAxes,
fontsize=12,
verticalalignment='top',
bbox=props)
plt.title(str(par), fontsize=20)
fig4.show()