-
Notifications
You must be signed in to change notification settings - Fork 0
/
heatmap.py
165 lines (133 loc) · 5.74 KB
/
heatmap.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
# @credit to https://github.com/drazenz/heatmap
from matplotlib import pyplot as plt
import pandas as pd
import seaborn as sns
import numpy as np
import torch
def heatmap(x, y, **kwargs):
if 'color' in kwargs:
color = kwargs['color']
else:
color = [1]*len(x)
if 'palette' in kwargs:
palette = kwargs['palette']
n_colors = len(palette)
else:
n_colors = 256 # Use 256 colors for the diverging color palette
palette = sns.color_palette("Blues", n_colors)
if 'color_range' in kwargs:
color_min, color_max = kwargs['color_range']
else:
color_min, color_max = min(color), max(color) # Range of values that will be mapped to the palette, i.e. min and max possible correlation
def value_to_color(val):
if color_min == color_max:
return palette[-1]
else:
val_position = float((val - color_min)) / (color_max - color_min) # position of value in the input range, relative to the length of the input range
val_position = min(max(val_position, 0), 1) # bound the position betwen 0 and 1
ind = int(val_position * (n_colors - 1)) # target index in the color palette
return palette[ind]
if 'size' in kwargs:
size = kwargs['size']
else:
size = [1]*len(x)
if 'size_range' in kwargs:
size_min, size_max = kwargs['size_range'][0], kwargs['size_range'][1]
else:
size_min, size_max = min(size), max(size)
size_scale = kwargs.get('size_scale', 500)
def value_to_size(val):
if size_min == size_max:
return 1 * size_scale
else:
val_position = (val - size_min) * 0.99 / (size_max - size_min) + 0.01 # position of value in the input range, relative to the length of the input range
val_position = min(max(val_position, 0), 1) # bound the position betwen 0 and 1
return val_position * size_scale
if 'x_order' in kwargs:
x_names = [t for t in kwargs['x_order']]
else:
x_names = [t for t in sorted(set([v for v in x]))]
x_to_num = {p[1]:p[0] for p in enumerate(x_names)}
if 'y_order' in kwargs:
y_names = [t for t in kwargs['y_order']]
else:
y_names = [t for t in sorted(set([v for v in y]))]
y_to_num = {p[1]:p[0] for p in enumerate(y_names)}
plot_grid = plt.GridSpec(1, 15, hspace=0.2, wspace=0.1) # Setup a 1x10 grid
ax = plt.subplot(plot_grid[:,:-1]) # Use the left 14/15ths of the grid for the main plot
marker = kwargs.get('marker', 's')
kwargs_pass_on = {k:v for k,v in kwargs.items() if k not in [
'color', 'palette', 'color_range', 'size', 'size_range', 'size_scale', 'marker', 'x_order', 'y_order'
]}
ax.scatter(
x=[x_to_num[v] for v in x],
y=[y_to_num[v] for v in y],
marker=marker,
s=[value_to_size(v) for v in size],
c=[value_to_color(v) for v in color],
**kwargs_pass_on
)
ax.set_xticks([v for k,v in x_to_num.items()])
ax.set_xticklabels([k for k in x_to_num], rotation=45, horizontalalignment='right')
ax.set_yticks([v for k,v in y_to_num.items()])
ax.set_yticklabels([k for k in y_to_num])
ax.grid(False, 'major')
ax.grid(True, 'minor')
ax.set_xticks([t + 0.5 for t in ax.get_xticks()], minor=True)
ax.set_yticks([t + 0.5 for t in ax.get_yticks()], minor=True)
ax.set_xlim([-0.5, max([v for v in x_to_num.values()]) + 0.5])
ax.set_ylim([-0.5, max([v for v in y_to_num.values()]) + 0.5])
ax.set_facecolor('#F1F1F1')
plt.title('Factor Correlations')
# Add color legend on the right side of the plot
if color_min < color_max:
ax = plt.subplot(plot_grid[:,-1]) # Use the rightmost column of the plot
col_x = [0]*len(palette) # Fixed x coordinate for the bars
bar_y=np.linspace(color_min, color_max, n_colors) # y coordinates for each of the n_colors bars
bar_height = bar_y[1] - bar_y[0]
ax.barh(
y=bar_y,
width=[5]*len(palette), # Make bars 5 units wide
left=col_x, # Make bars start at 0
height=bar_height,
color=palette,
linewidth=0
)
ax.set_xlim(1, 2) # Bars are going from 0 to 5, so lets crop the plot somewhere in the middle
ax.grid(False) # Hide grid
ax.set_facecolor('white') # Make background white
ax.set_xticks([]) # Remove horizontal ticks
ax.set_yticks(np.linspace(min(bar_y), max(bar_y), 3)) # Show vertical ticks for min, middle and max
ax.yaxis.tick_right() # Show vertical ticks on the right
def corrplot(data, size_scale=100, marker='s'):
corr = pd.melt(data.reset_index(), id_vars='index')
corr.columns = ['x', 'y', 'value']
colors = ["windows blue", "amber", "greyish", "faded green", "dusty purple"]
heatmap(
corr['x'], corr['y'],
color=corr['value'], color_range=[-1, 1],
palette=sns.diverging_palette(20, 220, n=256),
size=corr['value'].abs(),
marker=marker,
x_order=data.columns,
y_order=data.columns[::-1],
size_scale=size_scale
)
def norm_topics(numTopics, ell):
topk_similar_topics = torch.ones(ell.shape[0],ell.shape[0])
for k in range(numTopics):
topk_similar_topics[k] = torch.norm(ell[k:(k+1),:].repeat(numTopics,1)-ell,dim=1)
return topk_similar_topics
ell = torch.load('RUN_60_6_LayerDepth/ell.pt')
ellR_2 = norm_topics(ell.shape[0], ell.detach())
ell = ell.detach().cpu().numpy()
ellR = np.corrcoef(ell)
df = pd.DataFrame(ellR)
plt.figure(figsize=(20,20))
corrplot(df)
plt.savefig('topic_correlation_0_60_another.pdf')
ellR_2 = ellR_2.cpu().numpy()
df = pd.DataFrame(ellR_2)
plt.figure(figsize=(20,20))
corrplot(df)
plt.savefig('topic_correlation_0_60_norm.pdf')