-
Notifications
You must be signed in to change notification settings - Fork 3
/
plot.py
265 lines (225 loc) · 8.06 KB
/
plot.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
import csv
import numpy as np
import matplotlib as mpl
import matplotlib.pyplot as plt
from matplotlib.path import Path
from matplotlib.patches import PathPatch
from copy import deepcopy
def load_csv(filename):
with open(filename, "r") as f:
reader = csv.reader(f)
data_list = list(reader)
results = []
for i in range(1, len(data_list)):
result = []
for j in data_list[i]:
if "." in j or "e" in j:
try:
result.append(float(j))
except ValueError:
result.append(j)
else:
try:
result.append(int(j))
except ValueError:
result.append(j)
results.append(result)
return results, data_list[0]
def check_data(data, labels):
for i in range(len(data)):
assert len(data[i]) is len(labels), "data dimension (%d) does not " \
"match with labels (%d)" % (len(data[i]), len(labels))
def check_formatting(yattribute, labels):
if yattribute:
assert len(yattribute) is len(labels), "dimension (%d) does not " \
"match with labels (%d)" % (len(yattribute), len(labels))
else:
yattribute = [[]] * len(labels)
return yattribute
def set_ytype(ytype, data, colorbar):
for i in range(len(ytype)):
if not ytype[i]:
if type(data[0][i]) is str:
ytype[i] = "categorial"
else:
ytype[i] = "linear"
if colorbar:
assert ytype[len(ytype) - 1] == "linear", "colorbar axis needs to " \
"be linear"
return ytype
def set_ylabels(ylabels, data, ytype):
for i in range(len(ylabels)):
# Generate ylabels for string values
if not ylabels[i] and ytype[i] == "categorial":
ylabel = []
for j in range(len(data)):
if data[j][i] not in ylabel:
ylabel.append(data[j][i])
ylabel.sort()
if len(ylabel) == 1:
ylabel.append("")
ylabels[i] = ylabel
return ylabels
def replace_str_values(data, ytype, ylabels):
for i in range(len(ytype)):
if ytype[i] == "categorial":
for j in range(len(data)):
data[j][i] = ylabels[i].index(data[j][i])
return np.array(data).transpose()
def set_ylim(ylim, data):
for i in range(len(ylim)):
if not ylim[i]:
ylim[i] = [np.min(data[i, :]), np.max(data[i, :])]
if ylim[i][0] == ylim[i][1]:
ylim[i] = [ylim[i][0] * 0.95, ylim[i][1] * 1.05]
if ylim[i] == [0.0, 0.0]:
ylim[i] = [0.0, 1.0]
return ylim
def get_score(data, ylim):
ymin = ylim[len(ylim) - 1][0]
ymax = ylim[len(ylim) - 1][1]
score = (np.copy(data[len(ylim) - 1, :]) - ymin) / (ymax - ymin)
return score
# Rescale data of secondary y-axes to scale of first y-axis
def rescale_data(data, ytype, ylim):
min0 = ylim[0][0]
max0 = ylim[0][1]
scale = max0 - min0
for i in range(1, len(ylim)):
mini = ylim[i][0]
maxi = ylim[i][1]
if ytype[i] == "log":
logmin = np.log10(mini)
logmax = np.log10(maxi)
span = logmax - logmin
data[i, :] = ((np.log10(data[i, :]) - logmin) / span) * scale + min0
else:
data[i, :] = ((data[i, :] - mini) / (maxi - mini)) * scale + min0
return data
def get_path(data, i):
n = data.shape[0] # number of y-axes
verts = list(zip([x for x in np.linspace(0, n - 1, n * 3 - 2)],
np.repeat(data[:, i], 3)[1:-1]))
codes = [Path.MOVETO] + [Path.CURVE4 for _ in range(len(verts) - 1)]
path = Path(verts, codes)
return path
def pcp(data,
labels,
ytype=None,
ylim=None,
ylabels=None,
figsize=(10, 5),
rect=[0.125, 0.1, 0.75, 0.8],
curves=True,
alpha=1.0,
colorbar=True,
colorbar_width=0.02,
cmap=plt.get_cmap("inferno")
):
"""
Parallel Coordinates Plot
Parameters
----------
data: nested array
Inner arrays containing data for each curve.
labels: list
Labels for y-axes.
ytype: list, optional
Default "None" allows linear axes for numerical values and categorial
axes for data of type string. If ytype is passed, logarithmic axes are
also possible, e.g. ["categorial", "linear", "log", [], ...]. Vacant
fields must be filled with an empty list [].
ylim: list, optional
Custom min and max values for y-axes, e.g. [[0, 1], [], ...].
ylabels: list, optional (not recommended)
Only use this option if you want to print more categories than you have
in your dataset for categorial axes. You also have to set the right
ylim for this option to work correct.
figsize: (float, float), optional
Width, height in inches.
rect: array, optional
[left, bottom, width, height], defines the position of the figure on
the canvas.
curves: bool, optional
If True, B-spline curve is drawn.
alpha: float, optional
Alpha value for blending the curves.
colorbar: bool, optional
If True, colorbar is drawn.
colorbar_width: float, optional
Defines the width of the colorbar.
cmap: matplotlib.colors.Colormap, optional
Specify colors for colorbar.
Returns
-------
`~matplotlib.figure.Figure`
"""
[left, bottom, width, height] = rect
data = deepcopy(data)
# Check data
check_data(data, labels)
ytype = check_formatting(ytype, labels)
ylim = check_formatting(ylim, labels)
ylabels = check_formatting(ylabels, labels)
# Setup data
ytype = set_ytype(ytype, data, colorbar)
ylabels = set_ylabels(ylabels, data, ytype)
data = replace_str_values(data, ytype, ylabels)
ylim = set_ylim(ylim, data)
score = get_score(data, ylim)
data = rescale_data(data, ytype, ylim)
# Create figure
fig = plt.figure(figsize=figsize)
ax0 = fig.add_axes([left, bottom, width, height])
axes = [ax0] + [ax0.twinx() for i in range(data.shape[0] - 1)]
# Plot curves
for i in range(data.shape[1]):
if colorbar:
color = cmap(score[i])
else:
color = "blue"
if curves:
path = get_path(data, i)
patch = PathPatch(path, facecolor="None", lw=1.5, alpha=alpha,
edgecolor=color, clip_on=False)
ax0.add_patch(patch)
else:
ax0.plot(data[:, i], color=color, alpha=alpha, clip_on=False)
# Format x-axis
ax0.xaxis.tick_top()
ax0.xaxis.set_ticks_position("none")
ax0.set_xlim([0, data.shape[0] - 1])
ax0.set_xticks(range(data.shape[0]))
ax0.set_xticklabels(labels)
# Format y-axis
for i, ax in enumerate(axes):
ax.spines["left"].set_position(("axes", 1 / (len(labels) - 1) * i))
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["bottom"].set_visible(False)
ax.yaxis.set_ticks_position("left")
ax.set_ylim(ylim[i])
if ytype[i] == "log":
ax.set_yscale("log")
if ytype[i] == "categorial":
ax.set_yticks(range(len(ylabels[i])))
if ylabels[i]:
ax.set_yticklabels(ylabels[i])
if colorbar:
bar = fig.add_axes([left + width, bottom, colorbar_width, height])
norm = mpl.colors.Normalize(vmin=ylim[i][0], vmax=ylim[i][1])
mpl.colorbar.ColorbarBase(bar, cmap=cmap, norm=norm,
orientation="vertical")
bar.tick_params(size=0)
bar.set_yticklabels([])
return fig
if __name__ == "__main__":
# Minimal working example
results = [["ResNet", 0.0001, 4, 0.2],
["ResNet", 0.0003, 8, 1.0],
["DenseNet", 0.0005, 4, 0.65],
["DenseNet", 0.0007, 8, 0.45],
["DenseNet", 0.001, 2, 0.8]]
labels = ["Network", "Learning rate", "Batchsize", "F-Score"]
pcp(results, labels)
plt.show()