SCD/model.py at main · zeng-zhen/SCD · 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
import torch
import torch.nn as nn
import torch.nn.functional as F
# from fusion import Fusion
from gnn import GATLayer


class Net(nn.Module):
    def __init__(self, args):
        self.device = torch.device(
            ('cuda:%d' % (args.gpu)) if torch.cuda.is_available() else 'cpu')
        self.knowledge_dim = args.knowledge_n
        self.exer_n = args.exer_n
        self.emb_num = args.student_n
        self.stu_dim = self.knowledge_dim
        self.prednet_input_len = self.knowledge_dim
        self.prednet_len1, self.prednet_len2 = 512, 256
        super(Net, self).__init__()
        self.alphas = args.alphas
        self.alphae = args.alphae
        # network structure
        self.student_emb = nn.Embedding(self.emb_num, self.stu_dim)
        self.exercise_emb = nn.Embedding(self.exer_n, self.knowledge_dim)
        self.knowledge_emb = nn.Embedding(
            args.knowledge_n, self.knowledge_dim)

        self.gnet1 = GATLayer(
            self.knowledge_dim, self.knowledge_dim, self.knowledge_dim)
        self.gnet2 = GATLayer(
            self.knowledge_dim, self.knowledge_dim, self.knowledge_dim)

        self.prednet_full1 = nn.Linear(args.knowledge_n, args.knowledge_n, bias=False)

        self.prednet_full2 = nn.Linear(args.knowledge_n, args.knowledge_n, bias=False)

        self.prednet_full3 = nn.Linear(args.knowledge_n, args.knowledge_n)

        # initialization
        for name, param in self.named_parameters():
            if 'weight' in name:
                nn.init.xavier_normal_(param)

    def forward(self, stu_id, exer_id, kn_r, input_nodes, output_nodes, blocks, stu_index, exer_index, g1=None, g2=None):
        stu_emb = self.student_emb(input_nodes['stu']).to(self.device)
        exer_emb = self.exercise_emb(input_nodes['exer']).to(self.device)
        k_emb = self.knowledge_emb(input_nodes['k']).to(self.device)

        node_emb = {'stu': stu_emb, 'exer': exer_emb, 'k': k_emb}
        node_emb1 = self.gnet1(blocks[0], node_emb)
        node_emb2 = self.gnet2(blocks[1], node_emb1)

        exer_emb2 = node_emb2['exer']
        stu_emb2 = node_emb2['stu']

        stu_emb_g1 = self.student_emb(g1[0]['stu']).to(self.device)
        exer_emb_g1 = self.exercise_emb(g1[0]['exer']).to(self.device)
        k_emb_g1 = self.knowledge_emb(g1[0]['k']).to(self.device)
        node_emb_g1 = {'stu': stu_emb_g1, 'exer': exer_emb_g1, 'k': k_emb_g1}
        graph1_node_emb1 = self.gnet1(g1[2][0], node_emb_g1)
        graph1_node_emb2 = self.gnet2(g1[2][1], graph1_node_emb1)
        stu_emb2_g1 = graph1_node_emb2['stu']
        exer_emb2_g1 = graph1_node_emb2['exer']
        # e_map2
        stu_emb_g2 = self.student_emb(g2[0]['stu']).to(self.device)
        exer_emb_g2 = self.exercise_emb(g2[0]['exer']).to(self.device)
        k_emb_g2 = self.knowledge_emb(g2[0]['k']).to(self.device)
        node_emb_g2 = {'stu': stu_emb_g2, 'exer': exer_emb_g2, 'k': k_emb_g2}
        graph2_node_emb1 = self.gnet1(g2[2][0], node_emb_g2)
        graph2_node_emb2 = self.gnet2(g2[2][1], graph2_node_emb1)
        stu_emb2_g2 = graph2_node_emb2['stu']
        exer_emb2_g2 = graph2_node_emb2['exer']
        c_s_h1_loss = self.contrastive_loss(stu_emb2_g1, stu_emb2_g2)
        c_s_h2_loss = self.contrastive_loss(stu_emb2_g2, stu_emb2_g1)
        c_e_h1_loss = self.contrastive_loss(exer_emb2_g1, exer_emb2_g2)
        c_e_h2_loss = self.contrastive_loss(exer_emb2_g2, exer_emb2_g1)
        contrastive_loss = (self.alphas * (c_s_h1_loss + c_s_h2_loss) +
                                self.alphae * (c_e_h1_loss + c_e_h2_loss))

        batch_stu_emb = stu_emb2[stu_index]
        batch_exer_emb = exer_emb2[exer_index]

        preference = torch.sigmoid(self.prednet_full1(batch_stu_emb))
        diff = torch.sigmoid(self.prednet_full2(batch_exer_emb))
        o = torch.sigmoid(self.prednet_full3(preference - diff))
        sum_out = torch.sum(o * kn_r, dim=1)
        count_of_concept = torch.sum(kn_r, dim=1)
        output = sum_out / count_of_concept
        output = output.unsqueeze(1)

        return output, contrastive_loss

    def forward_test(self, stu_id, exer_id, kn_r, input_nodes, out_dict, blocks):
        stu_emb = self.student_emb(input_nodes['stu']).to(self.device)
        exer_emb = self.exercise_emb(input_nodes['exer']).to(self.device)
        k_emb = self.knowledge_emb(input_nodes['k']).to(self.device)

        node_emb = {'stu': stu_emb, 'exer': exer_emb, 'k': k_emb}
        # Fusion layer 1
        node_emb1 = self.gnet1(blocks[0], node_emb)
        # Fusion layer 2
        node_emb2 = self.gnet2(blocks[1], node_emb1)

        exer_emb2 = node_emb2['exer']
        stu_emb2 = node_emb2['stu']

        batch_stu_emb = stu_emb2.repeat(exer_emb2.shape[0], 1)
        batch_exer_emb = exer_emb2

        preference = torch.sigmoid(self.prednet_full1(batch_stu_emb))
        diff = torch.sigmoid(self.prednet_full2(batch_exer_emb))
        o = torch.sigmoid(self.prednet_full3(preference - diff))
        sum_out = torch.sum(o * kn_r, dim=1)
        count_of_concept = torch.sum(kn_r, dim=1)
        output = sum_out / count_of_concept
        output = output.unsqueeze(1)
        return output

    def apply_clipper(self):
        clipper = NoneNegClipper()
        self.prednet_full1.apply(clipper)
        self.prednet_full2.apply(clipper)
        self.prednet_full3.apply(clipper)


    def contrastive_loss(self, h1, h2):
        t = 0.5
        batch_size = h1.shape[0]
        negatives_mask = (~torch.eye(batch_size, batch_size,
                          dtype=bool)).to(self.device).float()
        z1 = F.normalize(h1, dim=1)
        z2 = F.normalize(h2, dim=1)
        similarity_matrix1 = F.cosine_similarity(
            z1.unsqueeze(1), z2.unsqueeze(0), dim=2)
        positives = torch.exp(torch.diag(similarity_matrix1) / t)
        negatives = negatives_mask * torch.exp(similarity_matrix1 / t)
        loss_partial = -torch.log(positives / (positives + torch.sum(negatives, dim=1)))
        loss = torch.sum(loss_partial) / batch_size
        return loss


class NoneNegClipper(object):
    def __init__(self):
        super(NoneNegClipper, self).__init__()

    def __call__(self, module):
        if hasattr(module, 'weight'):
            w = module.weight.data
            a = torch.relu(torch.neg(w))
            w.add_(a)