MASP/code/model.py at main · MattExpCode/MASP · 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
# -- coding: utf-8 --
import torch
import torch.nn as nn
import numpy as np
from torch.autograd import Variable
import torch.nn.functional as F


def graph_construction():
    graph = [[0, 1, 1, 1], [1, 0, 0, 0], [1, 0, 0, 1], [1, 0, 1, 0]]
    graph = [[0, 1, 0, 0, 0, 0], [1, 0, 0, 0, 0, 0], [0, 0, 0, 1, 1, 0], [0, 0, 1, 0, 0, 0], [0, 0, 1, 0, 0, 1],
             [0, 0, 0, 0, 1, 0]]
    # print(graph[0])
    graph = torch.LongTensor(graph)
    # print(graph)
    return graph


def neighbor_embedding(current_embed, neighbor_graph):
    sum_embedding = torch.mm(neighbor_graph, current_embed)  # (k,n) * (n,d) = (k, d)
    sum_neighbor = torch.sum(neighbor_graph, dim=1)  # (k)
    k = sum_neighbor.shape[0]
    flag_tensor = torch.ones_like(sum_neighbor)
    torch.where(sum_neighbor == 0, flag_tensor, sum_neighbor)
    aver_embedding = sum_embedding / sum_neighbor.view(k, -1)  # (k,d) ./ (k,d)

    return aver_embedding


class IAGR(nn.Module):
    def __int__(self, conf):
        self.path_length = conf.path_length
        self.fusion_group = nn.Linear(conf.fuse_dim, conf.embed_dim)
        self.fusion_item = nn.Linear(conf.fuse_dim, conf.embed_dim)
        self.sig_item = nn.Sigmoid()
        self.sig_group = nn.Sigmoid()
        self.W_1 = nn.Linear(conf.embed_dim * 2, conf.embed_dim)
        self.W_2 = nn.Linear(conf.embed_dim * 2, conf.embed_dim)
        self.W_3 = nn.Linear(conf.embed_dim * 2, conf.embed_dim)
        self.relu_1 = nn.LeakyReLU(inplace=False)
        self.relu_2 = nn.LeakyReLU(inplace=False)
        self.relu_3 = nn.LeakyReLU(inplace=False)
        # self.o_1 = neighbor_embedding()

    def forward(self, group_embed_init, item_embed_init, group_feat, item_feat, neighbor_graphs):
        concatenate_group = torch.cat([group_embed_init, group_feat], dim=1)
        concatenate_item = torch.cat([item_embed_init, item_feat], dim=1)
        group_embedding = self.sig_group(self.fusion_group(concatenate_group))
        item_embedding = self.sig_item(self.fusion_item(concatenate_item))

        group_embedding_list = []
        group_embedding_first = neighbor_embedding(group_embedding, neighbor_graphs[0])  # (k,d)
        concatenate_group_embedding_first = torch.cat([group_embedding, group_embedding_first], dim=1)
        group_embedding_first = self.relu_1(self.W_1(concatenate_group_embedding_first))  # (k,d)
        group_embedding_list.append(group_embedding_first)
        if self.path_length > 1:
            group_embedding_second = neighbor_embedding(group_embedding_first, neighbor_graphs[0])  # (k,d)
            concatenate_group_embedding_second = torch.cat([group_embedding_first, group_embedding_second], dim=1)
            group_embedding_second = self.relu_2(self.W_2(concatenate_group_embedding_second))
            group_embedding_list.append(group_embedding_second)

        if self.path_length > 2:
            group_embedding_third = neighbor_embedding(group_embedding_second, neighbor_graphs[0])  # (k,d)
            concatenate_group_embedding_third = torch.cat([group_embedding_second, group_embedding_third], dim=1)
            group_embedding_third = self.relu_3(self.W_2(concatenate_group_embedding_third))
            group_embedding_list.append(group_embedding_third)


def gene_neighbor_seed(seeds, graph, contained_nodes):
    list = [0, 1, 3]
    tmp = graph[seeds]
    [w, h] = tmp.shape
    id_list = []
    print(w, h)
    tmp = tmp.data.numpy()
    print(tmp)
    for i in range(w):
        cur_tmp = tmp[i]
        contained_node = contained_nodes[i]
        print(cur_tmp)
        id_ = np.where(cur_tmp == 1)
        print(id_[0], type(id_))
        A = set(contained_node)
        B = set(id_[0])

        print('neighbor: ', B - A)
        print('new_contain:', B | A)


def gene_neighbor(graph, contained_nodes):
    tmp = graph

    [w, h] = tmp.shape
    id_list = []
    # print(w, h)
    tmp = tmp.data.numpy()
    # print(tmp)
    graph_new = []
    contained_nodes_new = []
    for i in range(w):
        cur_tmp = tmp[i]
        contained_node = contained_nodes[i]
        # print(cur_tmp)
        id_ = np.where(cur_tmp != 0)
        # print(id_[0], type(id_))
        A = set(contained_node)
        B = set(id_[0])
        ex_nodes = list(B & A)
        neighbor = list(B - A)

        graph[i][ex_nodes] = 0
        # graph_new.append(tmp_neigh)
        tmp_contained = list(B | A)
        contained_nodes_new.append(tmp_contained)
        # print('neighbor: ', neighbor)
        # print('new_contain:', tmp_contained)

    # graph_new = torch.stack(graph_new, dim=0)
    print(contained_nodes_new)
    print(graph)
    # print(graph_new)
    return graph, contained_nodes_new


def construct_graph_order():
    G = graph_construction()
    nodes = []
    for v in range(6):
        nodes.append([v])
    print('first:\n', G)
    G_ = G.clone()
    first_neighbor, nodes = gene_neighbor(G_, nodes)
    second_G = torch.mm(G, G)
    print('second:\n', second_G)
    second_neighbor, nodes = gene_neighbor(second_G, nodes)
    third_G = torch.mm(second_G, G)
    print('third:\n', third_G)
    third_neighbor, nodes = gene_neighbor(third_G, nodes)
    return first_neighbor, second_neighbor, third_neighbor


def parameter_revised(tensor_test):
    tensor_test[[0, 1, 2]] = 0
    return tensor_test


if __name__ == '__main__':
    net = IAGR()
    torch.manual_seed(1024)
    embedding_func = nn.Embedding(2, 5)
    embedding_func = nn.Embedding(2000, 64)

    batch = embedding_func(torch.LongTensor([0, 1]))
    print(batch)

    A = [[1, 2, 3], [0, 0, 0]]
    B = [[1, 1, 1], [0, 0, 0]]
    a_ = torch.Tensor(A)
    a_1 = torch.Tensor(B)
    b = torch.cat([a_, a_1], dim=1)
    print(b, '\n', b.shape)
    c = torch.sum(b, dim=1)
    flag_tensor = torch.ones_like(c)
    c = torch.where(c == 0, flag_tensor, c)

    print(c, '\n', c.shape[0])
    e = b.t() * c

    # e = torch.unsqueeze(c, dim=1)
    # e = F.pad(input=e, pad=(0, 5), mode='replicate')
    print(b)
    print(e.t())
    f = b / c.view(2, -1)
    print(f)

    # diag = torch.diag(second_G)
    # diag_mat = torch.diag_embed(diag)
    # # print(second_G - diag_mat)
    # third_G = torch.mm(second_G - diag_mat, G)
    # diag = torch.diag(third_G)
    # diag_mat = torch.diag_embed(diag)
    # print(third_G - diag_mat)