/
util.py
69 lines (58 loc) · 2.99 KB
/
util.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
import random
import pickle
def generate_decvals(num, mean, sigma):
return [random.normalvariate(mean, sigma) for _ in range(num)]
def load_dist_config(distid):
"""Creates a configuration with differently distributed
decision values for positives and negatives, which will
result in differently shaped performance curves.
distid = 1: high initial precision (near-horizontal ROC curve)
distid = 2: low initial precision (near-vertical ROC curve)
distid = 3: standard performance, (circle-segment-like ROC curve)
"""
if distid == 1:
dist_config = {'mean_pos': 1.0, 'sigma_pos': 0.3,
'mean_neg': 0.0, 'sigma_neg': 1.0}
elif distid == 2:
dist_config = {'mean_pos': 2.0, 'sigma_pos': 2.0,
'mean_neg': 0.0, 'sigma_neg': 1.0}
else:
dist_config = {'mean_pos': 1.0, 'sigma_pos': 1.0,
'mean_neg': 0.0, 'sigma_neg': 1.0}
return dist_config
def simulate_data(distid, num_pos, num_neg, known_pos_frac, known_neg_frac):
dist_config = load_dist_config(distid)
generate_pos_class_decvals = lambda num: generate_decvals(num, dist_config['mean_pos'], dist_config['sigma_pos'])
generate_neg_class_decvals = lambda num: generate_decvals(num, dist_config['mean_neg'], dist_config['sigma_neg'])
decision_values = generate_pos_class_decvals(num_pos) + generate_neg_class_decvals(num_neg)
labels = [True] * num_pos + [False] * num_neg
true_labels, labels, beta = induce_partial_labels(labels,
known_pos_frac,
known_neg_frac)
return labels, true_labels, decision_values, beta
def induce_partial_labels(labs, known_pos_frac, known_neg_frac):
true_labels = list(labs)
labels = list(labs)
total_pos = float(sum(true_labels))
total_neg = float(len(true_labels) - total_pos)
labeled_pos = int(total_pos * known_pos_frac)
labeled_neg = int(total_neg * known_neg_frac)
pos_idx = [idx for idx, lab in enumerate(true_labels) if lab]
neg_idx = [idx for idx, lab in enumerate(true_labels) if not lab]
relabel_pos_idx = random.sample(pos_idx, int(total_pos - labeled_pos))
relabel_neg_idx = random.sample(neg_idx, int(total_neg - labeled_neg))
for i in relabel_pos_idx:
labels[i] = None
for i in relabel_neg_idx:
labels[i] = None
beta = float(true_labels.count(True) - labels.count(True)) / labels.count(None)
return true_labels, labels, beta
def load_dataset(choice, known_pos_frac, known_neg_frac):
with open('data.pkl', 'rb') as f:
data = pickle.load(f)
labels = list(data[choice]['labels'])
decision_values = list(data[choice]['decision_values'])
true_labels, labels, beta = induce_partial_labels(labels,
known_pos_frac,
known_neg_frac)
return labels, true_labels, decision_values, beta