FedLU/Code/sample_unlearn.py at main · nju-websoft/FedLU · GitHub

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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
import os
import math
import torch
import random
import numpy as np

seed = 1998111213
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
np.random.seed(seed)
random.seed(seed)

def read_triples(file_path):
    data = []
    with open(file_path,"r",encoding="utf-8") as fin:
        for line in fin.readlines():
            h,r,t = line.strip().split()
            data.append((h,r,t))
    return data

def write_triples(file_path,data):
    with open(file_path,"w",encoding="utf-8") as fout:
        for (h,r,t) in data:
            fout.write(h+"\t"+r+"\t"+t+"\n")

def random_select(data,frac=0.01):
    entity_count = dict()
    relation_count = dict()
    for (h,r,t) in data:
        if h not in entity_count:
            entity_count[h] = 0
        if r not in relation_count:
            relation_count[r] = 0
        if t not in entity_count:
            entity_count[t] = 0
        entity_count[h] += 1
        relation_count[r] += 1
        entity_count[t] += 1
    result_size = math.ceil(len(data)*frac)
    result = set()
    batch_size = math.ceil(result_size/10)
    while(len(result)<result_size):
        result_batch = random.sample(data,batch_size)
        for (h,r,t) in result_batch:
            if(entity_count[h]<=1 or relation_count[r]<=1 or entity_count[t]<=1):
                continue
            else:
                entity_count[h] -= 1
                relation_count[r] -= 1
                entity_count[t] -= 1
                result.add((h,r,t))
    unlearn = list(result)[:result_size]
    false = []
    for (h,r,t) in unlearn:
        t_f = random.sample(entity_count.keys(),1)[0]
        while(t_f==h or t_f==t or ((h,r,t_f) in data)):
            t_f = random.sample(entity_count.keys(),1)[0]
        false.append((h,r,t_f))
    return unlearn,false

def relation_centric_select(data,frac=0.01):
    entity_count = dict()
    relation_count = dict()
    relation2triple = dict()
    for (h,r,t) in data:
        if h not in entity_count:
            entity_count[h] = 0
        if r not in relation_count:
            relation_count[r] = 0
        if t not in entity_count:
            entity_count[t] = 0
        entity_count[h] += 1
        relation_count[r] += 1
        entity_count[t] += 1
        if r not in relation2triple:
            relation2triple[r] = []
        relation2triple[r].append((h,r,t))
    result_size = math.ceil(len(data)*frac)
    result = set()
    relation_selected = set()
    while(len(result)<result_size):
        relation_batch_size = math.ceil(frac*len(relation2triple.keys()))
        relation_batch = random.sample(relation2triple.keys(),relation_batch_size)
        for r in relation_batch:
            if r not in relation_selected:
                relation_selected.add(r)
                triple_batch_size = math.ceil(len(relation2triple[r])/10)
                triple_batch = random.sample(relation2triple[r],triple_batch_size)
                for (h,r,t) in triple_batch:
                    if(entity_count[h]<=1 or entity_count[t]<=1 or relation_count[r]<=1):
                        continue
                    else:
                        entity_count[h] -= 1
                        entity_count[t] -= 1
                        relation_count[r] -= 1
                        result.add((h,r,t))
                if len(result)>=result_size:
                    break
    print("number of selected identical relations %i"%(len(relation_selected)))
    unlearn = list(result)[:result_size]
    false = []
    for (h,r,t) in unlearn:
        t_f = random.sample(entity_count.keys(),1)[0]
        while(t_f==h or t_f==t or ((h,r,t_f) in data)):
            t_f = random.sample(entity_count.keys(),1)[0]
        false.append((h,r,t_f))
    return unlearn,false

def construct_unlearn(inpath,frac=0.01,type="random"):
    trainfile = os.path.join(inpath,"train.txt")
    unlearnfile = os.path.join(inpath,"unlearn.txt")
    falsefile = os.path.join(inpath,"false.txt")
    train = read_triples(trainfile)
    print("total number: %i"%(len(train)))
    if type=="random":
        unlearn,false = random_select(train,frac=frac)
    elif type=="relation":
        unlearn,false  = relation_centric_select(train,frac=frac)
    else:
        raise ValueError("type must be random/relation, default random")
    print("unlearn number: %i"%(len(unlearn)))
    print("false number: %i"%(len(false)))
    write_triples(unlearnfile,unlearn)

inpath="../Data/FB15k-237/C3FL/0"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C3FL/1"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C3FL/2"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C5FL/0"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C5FL/1"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C5FL/2"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C5FL/3"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C5FL/4"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/0"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/1"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/2"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/3"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/4"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/5"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/6"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/7"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/8"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/C10FL/9"
construct_unlearn(inpath,type="random")

inpath="../Data/FB15k-237/R3FL/0"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R3FL/1"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R3FL/2"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R5FL/0"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R5FL/1"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R5FL/2"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R5FL/3"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R5FL/4"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/0"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/1"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/2"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/3"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/4"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/5"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/6"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/7"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/8"
construct_unlearn(inpath,type="random")
inpath="../Data/FB15k-237/R10FL/9"
construct_unlearn(inpath,type="random")