forked from mjiang89/CrossSpot
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcrossspot-less-dense.py
183 lines (179 loc) · 5.65 KB
/
crossspot-less-dense.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
from copy import *
from math import *
from random import *
MAX_NUM_SEED = 100
k_data = 4
vec_n_local = [3,5,8,10]
vec_n_global = [100,100,100,100]
c_global = 10000
ground_truth = [set(range(0,vec_n_local[k])) for k in range(0,k_data)]
def prob_metric(lower_ns,lower_c,upper_ns,upper_c):
for k in range(0,k_data):
if lower_ns[k] == 0: return 0.0
lower_ns[k] = 1.0*lower_ns[k]
if upper_ns[k] == 0: return 0.0
upper_ns[k] = 1.0*upper_ns[k]
if lower_c == 0: return 0.0
lower_c = 1.0*lower_c
if upper_c == 0: return 0.0
upper_c = 1.0*upper_c
term1 = lower_c*(log(lower_c)-log(upper_c)-1.0)
term2 = 1.0
term3 = 0.0
for k in range(0,k_data):
term2 *= lower_ns[k]/upper_ns[k]
term3 += log(lower_ns[k])-log(upper_ns[k])
ret = term1+upper_c*term2-lower_c*term3
return ret
fw = open('result.csv','w')
fw.write('c_local,best_precision,best_recall,best_f1score,average_f1score\n')
for c_local in [3000,2500,2000,1500,1200,1000,950,900,850,800,750,700,650,600,550,500,450,400]:
best_accuracy = [0.0,0.0,0.0]
average_f1score = 0.0
# <begin> Generate ER-Poisson data.
pair2val = {}
for i in range(0,c_local):
pair = ''
for k in range(0,k_data):
p = randint(0,vec_n_local[k]-1)
pair += ','+str(p)
pair = pair[1:]
if not pair in pair2val:
pair2val[pair] = 0
pair2val[pair] += 1
for i in range(0,c_global-c_local):
pair = ''
for k in range(0,k_data):
p = randint(vec_n_local[k],vec_n_global[k]-1)
pair += ','+str(p)
pair = pair[1:]
if not pair in pair2val:
pair2val[pair] = 0
pair2val[pair] += 1
data,lineno = [],-1
item2lineno = [{} for k in range(0,k_data)]
for pair in pair2val:
lineno += 1
entry = [0 for k in range(0,k_data+1)]
arr = pair.split(',')
for k in range(0,k_data):
item = int(arr[k])
if not item in item2lineno[k]:
item2lineno[k][item] = set()
item2lineno[k][item].add(lineno)
entry[k] = item
entry[k_data] = pair2val[pair]
data.append(entry)
# --- Data ready: data [x0,x1,...x(k-1),val] and item2lineno <end>
# Generate random seed.
for seedno in range(0,MAX_NUM_SEED):
seed = [[set() for k in range(0,k_data)],[set() for k in range(0,k_data)],0,0.0]
for k in range(0,k_data):
num_item = randint(1,vec_n_global[k])
list_item = range(0,vec_n_global[k])
shuffle(list_item)
for j in range(0,num_item):
item = list_item[j]
seed[0][k].add(item)
# Item sets ==> Lineno sets ==> Count of block ==> Metric.
block = copy(seed)
for k in range(0,k_data):
for item in block[0][k]:
if item in item2lineno[k]:
block[1][k] = block[1][k] | item2lineno[k][item]
linenoset = block[1][0]
for k in range(1,k_data):
linenoset = linenoset & block[1][k]
block[2] = 0
for lineno in linenoset:
block[2] += data[lineno][k_data]
vec_n_block = [len(block[0][k]) for k in range(0,k_data)]
c_block = block[2]
block[3] = prob_metric(copy(vec_n_block),copy(c_block),copy(vec_n_global),copy(c_global))
# Local Search.
metric_old = block[3]
while True:
list_mode = range(0,k_data)
shuffle(list_mode)
for k_adjust in list_mode:
# Adjust mode [k_adjust].
# print 'Adjusting mode',k_adjust,'...'
linenoset = set()
FIRST_K = True
for k_fixed in range(0,k_data):
if k_fixed == k_adjust:
continue
if FIRST_K:
linenoset = copy(block[1][k_fixed])
FIRST_K = False
else:
linenoset = linenoset & copy(block[1][k_fixed])
item2count = {}
for lineno in linenoset:
item = data[lineno][k_adjust]
count = data[lineno][k_data]
if not item in item2count:
item2count[item] = 0
item2count[item] += count
vec_n_block = [len(block[0][k]) for k in range(0,k_data)]
sort_item2count = sorted(item2count.items(),key=lambda x:-x[1])
num_item = len(sort_item2count)
if num_item == 0:
continue
[item,c_block],n = sort_item2count[0],1
vec_n_block[k_adjust] = n
metric_best = prob_metric(copy(vec_n_block),copy(c_block),copy(vec_n_global),copy(c_global))
itemset = set([item])
for i in range(1,num_item):
[item,count] = sort_item2count[i]
n += 1
vec_n_block[k_adjust] = n
metric_curr = prob_metric(copy(vec_n_block),copy(c_block+count),copy(vec_n_global),copy(c_global))
if metric_curr <= metric_best:
break
metric_best = metric_curr
c_block += count
itemset.add(item)
if metric_best > block[3]:
block[0][k_adjust] = itemset
block[1][k_adjust] = set()
for item in itemset:
block[1][k_adjust] = block[1][k_adjust] | item2lineno[k_adjust][item]
block[2] = c_block
block[3] = metric_best
# print [len(block[0][k]) for k in range(0,k_data)],block[2],block[3]
if block[3] == metric_old:
break
metric_old = block[3]
# Evaluation.
prediction = copy(block)
for k in range(0,k_data):
prediction[0][k] = prediction[0][k] & ground_truth[k]
for k in range(0,k_data):
prediction[1][k].clear()
for item in prediction[0][k]:
if item in item2lineno[k]:
prediction[1][k] = prediction[1][k] | item2lineno[k][item]
linenoset = prediction[1][0]
for k in range(1,k_data):
linenoset = linenoset & prediction[1][k]
hits = 0
for lineno in linenoset:
hits += data[lineno][k_data]
precision = 0.0
if prediction[2] > 0:
precision = 1.0*hits/prediction[2]
recall = 1.0*hits/c_local
f1score = 0.0
if precision+recall > 0:
f1score = 2*precision*recall/(precision+recall)
# print precision,recall,f1score
if f1score >= best_accuracy[2]:
best_accuracy = [precision,recall,f1score]
average_f1score += f1score
average_f1score /= MAX_NUM_SEED
s = str(c_local)+','+str(best_accuracy[0])+','+str(best_accuracy[1]) \
+','+str(best_accuracy[2])+','+str(average_f1score)
fw.write(s+'\n')
print s
fw.close()