-
Notifications
You must be signed in to change notification settings - Fork 1
/
evaluation.py
167 lines (137 loc) · 5.22 KB
/
evaluation.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
# !/usr/bin/env python
# -*- coding: utf8 -*-
import scipy.special
import numpy as np
import networkx as nx
from tabulate import tabulate
from sklearn.metrics import f1_score,adjusted_rand_score,normalized_mutual_info_score
# Get precicion
def getPrecision(mat, k, s, total):
sum_k = 0
for i in range(k):
max_s = 0
for j in range(s):
if mat[i][j] > max_s:
max_s = mat[i][j]
sum_k += max_s
return sum_k/total
# Get recall
def getRecall(mat, k, s, total, unclassified):
sum_s = 0
for i in range(s):
max_k = 0
for j in range(k):
if mat[j][i] > max_k:
max_k = mat[j][i]
sum_s += max_k
return sum_s/(total+unclassified)
# Get ARI
def getARI(mat, k, s, N):
t1 = 0
for i in range(k):
sum_k = 0
for j in range(s):
sum_k += mat[i][j]
t1 += scipy.special.binom(sum_k, 2)
t2 = 0
for i in range(s):
sum_s = 0
for j in range(k):
sum_s += mat[j][i]
t2 += scipy.special.binom(sum_s, 2)
t3 = t1*t2/scipy.special.binom(N, 2)
t = 0
for i in range(k):
for j in range(s):
t += scipy.special.binom(mat[i][j], 2)
ari = (t-t3)/((t1+t2)/2-t3)
return ari
# Get F1-score
def getF1(prec, recall):
if prec == 0.0 or recall == 0.0:
return 0.0
else:
return 2*prec*recall/(prec+recall)
def validate_performance(true_labels, pred_labels):
ReMap = {j:i for i,j in enumerate(np.unique([v for k,v in pred_labels.items()]))}
ReMap_true = {j:i for i,j in enumerate(np.unique([v for k,v in true_labels.items()]))}
# print(len(ReMap))
n_true_labels = len(np.unique([v for k,v in true_labels.items()]))
n_pred_labels = len(np.unique([v for k,v in pred_labels.items()]))
# print(np.unique([v for k,v in pred_labels.items()]))
# print(n_true_labels, n_pred_labels)
ground_truth_count = len(true_labels)
# print(n_true_labels, n_pred_labels)
total_binned = 0
bins_species = [[0 for x in range(n_true_labels)] for y in range(n_pred_labels)]
for i in pred_labels:
if i in true_labels:
total_binned += 1
bins_species[ReMap[pred_labels[i]]][ReMap_true[true_labels[i]]] += 1
my_precision = getPrecision(bins_species, n_pred_labels, n_true_labels, total_binned)
my_recall = getRecall(bins_species, n_pred_labels, n_true_labels, total_binned, (ground_truth_count-total_binned))
my_ari = getARI(bins_species, n_pred_labels, n_true_labels, total_binned)
my_f1 = getF1(my_precision, my_recall)
# print("### Evaluation:")
# print("### Precision = %0.4f Recall = %0.4f F1 = %0.4f ARI = %0.4f" % (my_precision, my_recall, my_f1, my_ari))
return my_precision, my_recall, my_ari, my_f1
from collections import defaultdict
def list_duplicates(seq):
tally = defaultdict(list)
for i,item in enumerate(seq):
tally[item].append(i)
return (locs for key,locs in tally.items() if len(locs)>1)
def Clustering(embeds, true_labels, constraints, Gx, n_clusters=5):
p_all, r_all, ari_all, f1_all = [],[],[],[]
for t in range(3):
from sklearn.cluster import KMeans
estimator = KMeans(n_clusters=n_clusters)
estimator.fit(embeds)
pred_labels_ = estimator.labels_
pred_labels = {i:j for i,j in enumerate(pred_labels_)}
p, r, ari, f1 = validate_performance(true_labels, pred_labels)
p_all.append(p)
r_all.append(r)
ari_all.append(ari)
f1_all.append(f1)
avg_p,avg_r,avg_ari,avg_f1 = np.mean(p_all),np.mean(r_all),np.mean(ari_all),np.mean(f1_all)
std_p,std_r,std_ari,std_f1 = np.std(p_all),np.std(r_all),np.std(ari_all),np.std(f1_all)
print()
print ("### Average (over trials): Precision = %0.4f(%0.4f) Recall = %0.4f(%0.4f) F1 = %0.4f(%0.4f) ARI = %0.4f(%0.4f)"
% (avg_p,std_p,avg_r,std_r,avg_f1,std_f1,avg_ari,std_ari))
return pred_labels
def Cluster(embeds, true_labels, n_clusters=5):
Mf1_all,mf1_all,nmi_all,ari_all = [],[],[],[]
for t in range(5):
from sklearn.cluster import KMeans
estimator = KMeans(n_clusters=n_clusters)
estimator.fit(embeds)
pred_labels_ = estimator.labels_
pred_labels = {i:j for i,j in enumerate(pred_labels_)}
true_idx = [key for key,val in true_labels.items()]
true_lbls = [val for key,val in true_labels.items()]
pred_lbls = [pred_labels[idx] for idx in true_idx]
Mf1 = f1_score(true_lbls, pred_lbls, average='macro')
Mf1_all.append(Mf1)
mf1 = f1_score(true_lbls, pred_lbls, average='micro')
mf1_all.append(mf1)
ari = adjusted_rand_score(true_lbls, pred_lbls)
ari_all.append(ari)
nmi = normalized_mutual_info_score(true_lbls, pred_lbls)
nmi_all.append(nmi)
avg_Mf1,avg_mf1,avg_ari,avg_nmi = np.mean(Mf1_all),np.mean(mf1_all),np.mean(ari_all),np.mean(nmi_all)
std_Mf1,std_mf1,std_ari,std_nmi = np.std(Mf1_all),np.std(mf1_all),np.std(ari_all),np.std(nmi_all)
print()
print ("### Average (over trials): macro-F1 = %0.4f(%0.4f), micro-F1 = %0.4f(%0.4f), ARI = %0.4f(%0.4f), NMI = %0.4f(%0.4f)"
% (avg_Mf1,std_Mf1,avg_mf1,std_mf1,avg_ari,std_ari,avg_nmi,std_nmi))
def validate_ARI_NMI(true_labels, pred_labels):
true_idx = [key for key,val in true_labels.items()]
true_lbls = [val for key,val in true_labels.items()]
pred_lbls = [pred_labels[idx] for idx in true_idx]
ari = adjusted_rand_score(true_lbls, pred_lbls)
nmi = normalized_mutual_info_score(true_lbls, pred_lbls)
return ari, nmi
def evaluate_performance(true_labels, pred_labels):
ari, nmi = validate_ARI_NMI(true_labels, pred_labels)
p, r, _, f1 = validate_performance(true_labels, pred_labels)
return p,r,f1,ari,nmi