-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathtsne-plots.py
135 lines (109 loc) · 4.71 KB
/
tsne-plots.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
import pandas as pd
import numpy as np
from os.path import basename, splitext
import glob
from sklearn.manifold import TSNE
import matplotlib.pyplot as plt
from matplotlib.pyplot import figure
from matplotlib.collections import PathCollection
from matplotlib.legend_handler import HandlerPathCollection, HandlerLine2D
import seaborn as sns
import os
sns.set_style("whitegrid")
def update(handle, orig):
handle.update_from(orig)
handle.set_alpha(1)
figure(figsize=(8, 6), dpi=80)
markers = ['x', '+', 'o', '*', 'x', '+']
emotions = ["joi", "col","pla","tri"]
emotion_english =['Elation','Anger','Pleasure','Sad']
colors = ['#695167','#B85B4F', '#9E8F8D', '#697848', 'black']
colorname= ['green','gray','purple','red']
home = 'embeddings/train8k256/train-embeddings/'
proto_files = glob.glob(home + '*.txt')
augs = ['wavegan','source']
augs_true = ['WaveGAN','Source']
# tsne hps
lr = 20
perp = 90
instances = 2
def point_distance(point1, point2, metric="EUCLIDEAN"):
if metric == "EUCLIDEAN":
return np.sqrt(np.sum((point1 - point2)**2))
def remvove_doubles(points):
new_points = []
for point in points:
min_dist = np.inf
for i in range(len(new_points)):
dist = point_distance(point, new_points[i])
if dist < min_dist:
min_dist = dist
if min_dist > 1e-10:
new_points.append(point)
new_points = np.array(new_points)
return new_points
for interp in range(0,instances):
for num_aug in range(len(augs)):
print(augs[num_aug])
aug_type = augs[num_aug]
home = 'embeddings/train8k256/test-train-embeddings/' + aug_type +'/'
embedding_files = glob.glob(home + '*.txt')
augmentation_techniques = []
result_dir = "tsne-plot/"
os.makedirs(result_dir, exist_ok=True)
prototypes = []
all_embeddings = []
all_num_embeddings = []
for num_emotion in range(len(emotions)):
emotion = emotions[num_emotion]
for embedding_file in embedding_files:
if not emotion in basename(embedding_file):
continue
augmentation_technique = "_".join(basename(embedding_file).split("_")[1:3])
augmentation_techniques.append(augmentation_technique)
proto = np.genfromtxt(proto_files[num_emotion])
proto = remvove_doubles(proto)
embeddings = np.genfromtxt(embedding_file)
proto = remvove_doubles(embeddings)
all_embeddings.append(embeddings)
all_num_embeddings.append(len(embeddings))
prototype = embeddings.mean(axis=0)[np.newaxis,...]
prototypes.append(prototype)
all_embeddings.append(prototype)
distance_matrix = np.empty((len(prototypes), len(prototypes)))
for i in range(len(prototypes)):
for j in range(len(prototypes)):
distance_matrix[i,j] = np.sqrt(np.sum((prototypes[i] - prototypes[j])**2))
print("----------------------------------------------------------------------------")
print(emotion)
print(augmentation_techniques)
all_embeddings = np.vstack(all_embeddings)
tsne_embeddings = TSNE(2,learning_rate=lr,verbose=0, perplexity=perp, n_iter=450).fit_transform(all_embeddings)
all_prototypes = np.vstack(prototypes)
tsne_prototypes = TSNE(2,learning_rate=lr,verbose=0, perplexity=perp, n_iter=450).fit_transform(all_prototypes)
current_count = 0
for num_emotion in range(len(emotions)):
emotion = emotions[num_emotion]
print(emotion_english[num_emotion])
for step in range(int(len(all_num_embeddings)/len(emotions))):
if step == 0:
pass
print("Step: {},".format(step) + emotion + ", " + augmentation_techniques[step])
num_embeddings = all_num_embeddings[step + num_emotion*int(len(all_num_embeddings)/len(emotions))]
embeddings = tsne_embeddings[current_count:current_count + num_embeddings]
prototype = tsne_embeddings[current_count + num_embeddings]
if aug_type == 'source':
plt.scatter(prototype[0], prototype[1], marker='o', s=500, color=colors[num_emotion], edgecolors='w', alpha=1)
else:
plt.scatter(prototype[0], prototype[1], marker=markers[num_aug], s=100, color=colors[num_emotion], edgecolors='w', alpha=0.1)
# ~ if num_emotion == 0:
# ~ plt.scatter(embeddings[:, 0], embeddings[:, 1],marker=markers[num_aug], s=80, color=colors[num_emotion], alpha=0.6)
# ~ else:
plt.scatter(embeddings[:, 0], embeddings[:, 1], marker=markers[num_aug],s=40, color=colors[num_emotion], alpha=0.6,label = emotion_english[num_emotion] + ' ' + augs_true[num_aug])
current_count += num_embeddings + 1
plt.legend(handler_map={PathCollection : HandlerPathCollection(update_func= update),
plt.Line2D : HandlerLine2D(update_func = update)})
plt.xlim([-30, 30])
plt.ylim([-30, 30])
plt.savefig(f'{result_dir}{interp}_{augs[num_aug]}_{emotion_english[0]}lr-{lr}-p-{perp}.pdf')
plt.clf()