def plot_weights(model_data):
category_list = model_data[0][0]
feature_list = model_data[0][1]
hidden_size = model_data[1][2]
h_x = model_data[2][0]
y_h = model_data[2][1]
h_labels = ['H0']
for i in range(hidden_size):
h_labels.append("H" + str(i + 1))
h1 = heatmapcluster.heatmapcluster(y_h.transpose(),
h_labels,
category_list,
label_fontsize=12,
xlabel_rotation=0,
cmap=plt.cm.coolwarm,
show_colorbar=True,
colorbar_pad=2,
top_dendrogram=True)
h2 = heatmapcluster.heatmapcluster(h_x,
h_labels,
feature_list,
label_fontsize=12,
xlabel_rotation=90,
cmap=plt.cm.coolwarm,
show_colorbar=True,
colorbar_pad=2,
top_dendrogram=True)
def plot_hierarchical_cluster(the_dataset, svd, similarity):
print("\n Plotting Hierarchical Cluster Diagram")
ok_to_go = False
try:
import heatmapcluster
ok_to_go = True
except:
print(
"Cannot import heatmapcluster. try running 'pip3 install heatmapcluster'"
)
if svd:
if the_dataset.svd_dimensions is None:
print(
"ERROR: Cannot make SVD heat map without specifying SVD dimensions"
)
sys.exit()
else:
data_matrix = the_dataset.svd_row_features[:, :the_dataset.
svd_dimensions]
else:
data_matrix = the_dataset.feature_matrix
if ok_to_go:
y_labels = the_dataset.word_list
if similarity is not None:
x_labels = the_dataset.word_list
data = np.zeros([the_dataset.num_words, the_dataset.num_words])
for i in range(the_dataset.num_words):
a = the_dataset.feature_matrix[i, :]
for j in range(the_dataset.num_words):
b = the_dataset.feature_matrix[j, :]
data[i, j] = np.dot(
a, b) / (np.linalg.norm(a) * np.linalg.norm(b))
else:
x_labels = the_dataset.feature_list
data = data_matrix
print(data_matrix.shape)
h = heatmapcluster.heatmapcluster(data,
y_labels,
x_labels,
label_fontsize=6,
xlabel_rotation=90,
cmap=plt.cm.coolwarm,
show_colorbar=True,
colorbar_pad=2,
top_dendrogram=True)
plt.show()
def calcCorMatrix(m, fn):
cor = [[0 for x in xrange(len(m))] for y in xrange(len(m))]
for i in xrange(len(m)):
for j in xrange(len(m)):
if m[i][0] != None and m[j][0] != None:
cor[i][j] = pearsonr(m[i], m[j])[0]
else:
cor[i][j] = None
fn = fn + ".csv"
f = open(fn, 'w')
strcor = str(cor)[1:-1].replace("], ", "\n").replace("[",
"").replace("]", "")
f.write(strcor)
f.close()
sp = ["MR", "R", "R2", "O", "I", "I2R2", "MY", "Y", "YIex"]
ch = ["dPK", "dPKT", "dHRT", "dHFT"]
names = list()
for x in xrange(len(sp)):
for y in xrange(len(ch)):
names.append(sp[x] + ' ' + ch[y])
names.remove("MR dPKT")
names.remove("MR dHRT")
print(names)
# names = list(reversed(names))
data = np.genfromtxt(fn, delimiter=",")
clustersize = 12
h = heatmapcluster(data,
names,
names,
num_row_clusters=clustersize,
num_col_clusters=clustersize,
label_fontsize=9,
xlabel_rotation=-75,
cmap=plt.cm.coolwarm,
show_colorbar=True,
top_dendrogram=True)
plt.savefig("cor_heatmap.png")
def plot_weight_heat_map(self):
ok_to_go = False
try:
import heatmapcluster
ok_to_go = True
except:
print(
"Cannot import heatmapcluster. try running 'pip install heatmapcluster'"
)
if ok_to_go:
h = heatmapcluster.heatmapcluster(self.y_x_weights.transpose(),
self.dataset.feature_list,
self.dataset.category_list,
label_fontsize=12,
xlabel_rotation=0,
cmap=plt.cm.coolwarm,
show_colorbar=True,
colorbar_pad=2,
top_dendrogram=True)
def plot_hierarchical_cluster(self, similarity=False):
print("\n Plotting Hierarchical Cluster Diagram")
ok_to_go = False
try:
import heatmapcluster
ok_to_go = True
except:
print(
"Cannot import heatmapcluster. try running 'pip install heatmapcluster'"
)
if ok_to_go:
x_labels = self.word_list
y_labels = self.word_list
if similarity:
data = np.zeros([self.num_words, self.num_words])
for i in range(self.num_words):
a = self.feature_matrix[i, :]
for j in range(self.num_words):
b = self.feature_matrix[j, :]
data[i, j] = np.dot(
a, b) / (np.linalg.norm(a) * np.linalg.norm(b))
else:
data = self.feature_matrix.transpose()
x_labels = self.word_list
y_labels = self.feature_list
h = heatmapcluster.heatmapcluster(data,
y_labels,
x_labels,
label_fontsize=6,
xlabel_rotation=90,
cmap=plt.cm.coolwarm,
show_colorbar=True,
colorbar_pad=2,
top_dendrogram=True)
if seed is not None:
np.random.seed(seed)
s = np.random.gamma([7, 6, 5], [6, 8, 6], size=(size[1], 3)).T
i = np.random.choice(range(len(s)), size=size[0])
x = s[i]
t = np.random.gamma([8, 5, 6], [3, 3, 2.1], size=(size[0], 3)).T
j = np.random.choice(range(len(t)), size=size[1])
x += 1.1*t[j].T
x += 2*np.random.randn(*size)
row_labels = [('R%02d' % k) for k in range(x.shape[0])]
col_labels = [('C%02d' % k) for k in range(x.shape[1])]
return x, row_labels, col_labels
x, row_labels, col_labels = make_data(size=(64, 48), seed=123)
h = heatmapcluster(x, row_labels, col_labels,
num_row_clusters=3, num_col_clusters=0,
label_fontsize=6,
xlabel_rotation=-75,
cmap=plt.cm.coolwarm,
show_colorbar=True,
top_dendrogram=True)
plt.show()
def main(argv):
try:
opts, args = getopt.getopt(argv, "", ["file="])
except getopt.GetoptError:
raise NameError("Invalid arguments. Call string should look like:\nportfolio_heatmap.py --file <path-to-folder>")
for opt, arg in opts:
if opt in ("--file"):
file = arg
else:
print ("Option " + opt + " is not recognized")
print ("permutation_folds.py --data-file <path-to-file> --outputdir <path-to-output-dir> --seed_range <begin,end>")
sys.exit(2)
maxValue=10000 #Cutoff??
with open(file, 'rU') as csvfile:
lines = csv.DictReader(csvfile)
values = []
vbsValues = []
for line in lines:
row = []
vbsRow = []
minPerformance = 100000000
bestSolver = ''
for col in line:
if (float(line[col]) < minPerformance):
minPerformance = float(line[col])
bestSolver = col
row.append(float(line[col]))
for col in line:
if col == bestSolver:
vbsRow.append(line[bestSolver])
else:
vbsRow.append(maxValue)
vbsValues.append(vbsRow)
values.append(row)
x_labels = list(xrange(len(values)))
y_labels = list(xrange(len(values[0])))
# Maybe you want to normalize each row???
'''
log_transform = False
if log_transform:
mypyutils.pyplot.plotHeatMap(values,x_labels, y_labels,colormap=plt.cm.get_cmap('RdBu'),norm=matplotlib.colors.LogNorm())
else:
mypyutils.pyplot.plotHeatMap(values,x_labels, y_labels,colormap=plt.cm.get_cmap('RdBu'))
plt.xlabel("Hydra Iterations")
plt.ylabel("Instances")
plt.title("Hydra Portfolio Performance Heatmap")
ax = plt.gca()
cbar = plt.colorbar(orientation='vertical',pad=0.025,shrink=0.75,fraction=0.05)
cbar.set_label('Performance', rotation=270,labelpad=15)
cbar.ax.set_aspect(20)
plt.show()
plt.clf()
log_transform = False
if log_transform:
mypyutils.pyplot.plotHeatMap(vbsValues,x_labels, y_labels,colormap=plt.cm.get_cmap('RdBu'),norm=matplotlib.colors.LogNorm())
else:
mypyutils.pyplot.plotHeatMap(vbsValues,x_labels, y_labels,colormap=plt.cm.get_cmap('RdBu'))
plt.xlabel("Hydra Iterations")
plt.ylabel("Instances")
plt.title("Hydra Portfolio Performance Heatmap")
ax = plt.gca()
cbar = plt.colorbar(orientation='vertical',pad=0.025,shrink=0.75,fraction=0.05)
cbar.set_label('Performance', rotation=270,labelpad=15)
cbar.ax.set_aspect(20)
'''
np_values = np.array(values)
h = heatmapcluster(np_values, x_labels, y_labels,
num_row_clusters=3, num_col_clusters=0,
label_fontsize=6,
xlabel_rotation=-75,
cmap=plt.cm.coolwarm,
show_colorbar=True,
top_dendrogram=False)
plt.show()
'''
j = np.random.choice(range(len(t)), size=size[1])
x += 1.1 * t[j].T
x += 2 * np.random.randn(*size)
row_labels = [('R%02d' % k) for k in range(x.shape[0])]
col_labels = [('C%02d' % k) for k in range(x.shape[1])]
return x, row_labels, col_labels
x, row_labels, col_labels = make_data(size=(64, 48), seed=123)
h = heatmapcluster(
x,
row_labels,
col_labels,
num_row_clusters=3,
num_col_clusters=0,
label_fontsize=6,
xlabel_rotation=-75,
cmap=plt.cm.coolwarm,
show_colorbar=True,
top_dendrogram=True,
row_linkage=lambda x: linkage(x, method='average', metric='correlation'),
col_linkage=lambda x: linkage(x.T, method='average', metric='correlation'),
histogram=True)
plt.show()
t = np.random.gamma([8, 5, 6], [3, 3, 2.1], size=(size[0], 3)).T
j = np.random.choice(range(len(t)), size=size[1])
x += 1.1*t[j].T
x += 2*np.random.randn(*size)
row_labels = [('R%02d' % k) for k in range(x.shape[0])]
col_labels = [('C%02d' % k) for k in range(x.shape[1])]
return x, row_labels, col_labels
x, row_labels, col_labels = make_data(size=(64, 48), seed=123)
h = heatmapcluster(x, row_labels, col_labels,
num_row_clusters=3, num_col_clusters=0,
label_fontsize=6,
xlabel_rotation=-75,
cmap=plt.cm.coolwarm,
show_colorbar=True,
top_dendrogram=True,
row_linkage=lambda x: linkage(x, method='average',
metric='correlation'),
col_linkage=lambda x: linkage(x.T, method='average',
metric='correlation'),
histogram=True)
plt.show()