File size: 5,671 Bytes
e1ccef5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import torch.nn as nn
import torch.nn.functional as F


class IEBlock(nn.Module):
    def __init__(self, input_dim, hid_dim, output_dim, num_node):
        super(IEBlock, self).__init__()

        self.input_dim = input_dim
        self.hid_dim = hid_dim
        self.output_dim = output_dim
        self.num_node = num_node

        self._build()

    def _build(self):
        self.spatial_proj = nn.Sequential(
            nn.Linear(self.input_dim, self.hid_dim),
            nn.LeakyReLU(),
            nn.Linear(self.hid_dim, self.hid_dim // 4)
        )

        self.channel_proj = nn.Linear(self.num_node, self.num_node)
        torch.nn.init.eye_(self.channel_proj.weight)

        self.output_proj = nn.Linear(self.hid_dim // 4, self.output_dim)

    def forward(self, x):
        x = self.spatial_proj(x.permute(0, 2, 1))
        x = x.permute(0, 2, 1) + self.channel_proj(x.permute(0, 2, 1))
        x = self.output_proj(x.permute(0, 2, 1))

        x = x.permute(0, 2, 1)

        return x


class Model(nn.Module):
    """
    Paper link: https://arxiv.org/abs/2207.01186
    """

    def __init__(self, configs, chunk_size=24):
        """
        chunk_size: int, reshape T into [num_chunks, chunk_size]
        """
        super(Model, self).__init__()
        self.task_name = configs.task_name
        self.seq_len = configs.seq_len
        if self.task_name == 'classification' or self.task_name == 'anomaly_detection' or self.task_name == 'imputation':
            self.pred_len = configs.seq_len
        else:
            self.pred_len = configs.pred_len

        if configs.task_name == 'long_term_forecast' or configs.task_name == 'short_term_forecast':
            self.chunk_size = min(configs.pred_len, configs.seq_len, chunk_size)
        else:
            self.chunk_size = min(configs.seq_len, chunk_size)
        # assert (self.seq_len % self.chunk_size == 0)
        if self.seq_len % self.chunk_size != 0:
            self.seq_len += (self.chunk_size - self.seq_len % self.chunk_size)  # padding in order to ensure complete division
        self.num_chunks = self.seq_len // self.chunk_size

        self.d_model = configs.d_model
        self.enc_in = configs.enc_in
        self.dropout = configs.dropout
        if self.task_name == 'classification':
            self.act = F.gelu
            self.dropout = nn.Dropout(configs.dropout)
            self.projection = nn.Linear(configs.enc_in * configs.seq_len, configs.num_class)
        self._build()

    def _build(self):
        self.layer_1 = IEBlock(
            input_dim=self.chunk_size,
            hid_dim=self.d_model // 4,
            output_dim=self.d_model // 4,
            num_node=self.num_chunks
        )

        self.chunk_proj_1 = nn.Linear(self.num_chunks, 1)

        self.layer_2 = IEBlock(
            input_dim=self.chunk_size,
            hid_dim=self.d_model // 4,
            output_dim=self.d_model // 4,
            num_node=self.num_chunks
        )

        self.chunk_proj_2 = nn.Linear(self.num_chunks, 1)

        self.layer_3 = IEBlock(
            input_dim=self.d_model // 2,
            hid_dim=self.d_model // 2,
            output_dim=self.pred_len,
            num_node=self.enc_in
        )

        self.ar = nn.Linear(self.seq_len, self.pred_len)

    def encoder(self, x):
        B, T, N = x.size()

        highway = self.ar(x.permute(0, 2, 1))
        highway = highway.permute(0, 2, 1)

        # continuous sampling
        x1 = x.reshape(B, self.num_chunks, self.chunk_size, N)
        x1 = x1.permute(0, 3, 2, 1)
        x1 = x1.reshape(-1, self.chunk_size, self.num_chunks)
        x1 = self.layer_1(x1)
        x1 = self.chunk_proj_1(x1).squeeze(dim=-1)

        # interval sampling
        x2 = x.reshape(B, self.chunk_size, self.num_chunks, N)
        x2 = x2.permute(0, 3, 1, 2)
        x2 = x2.reshape(-1, self.chunk_size, self.num_chunks)
        x2 = self.layer_2(x2)
        x2 = self.chunk_proj_2(x2).squeeze(dim=-1)

        x3 = torch.cat([x1, x2], dim=-1)

        x3 = x3.reshape(B, N, -1)
        x3 = x3.permute(0, 2, 1)

        out = self.layer_3(x3)

        out = out + highway
        return out

    def forecast(self, x_enc, x_mark_enc, x_dec, x_mark_dec):
        return self.encoder(x_enc)

    def imputation(self, x_enc, x_mark_enc, x_dec, x_mark_dec, mask):
        return self.encoder(x_enc)

    def anomaly_detection(self, x_enc):
        return self.encoder(x_enc)

    def classification(self, x_enc, x_mark_enc):
        # padding
        x_enc = torch.cat([x_enc, torch.zeros((x_enc.shape[0], self.seq_len-x_enc.shape[1], x_enc.shape[2])).to(x_enc.device)], dim=1)

        enc_out = self.encoder(x_enc)

        # Output
        output = enc_out.reshape(enc_out.shape[0], -1)  # (batch_size, seq_length * d_model)
        output = self.projection(output)  # (batch_size, num_classes)
        return output

    def forward(self, x_enc, x_mark_enc, x_dec, x_mark_dec, mask=None):
        if self.task_name == 'long_term_forecast' or self.task_name == 'short_term_forecast':
            dec_out = self.forecast(x_enc, x_mark_enc, x_dec, x_mark_dec)
            return dec_out[:, -self.pred_len:, :]  # [B, L, D]
        if self.task_name == 'imputation':
            dec_out = self.imputation(x_enc, x_mark_enc, x_dec, x_mark_dec, mask)
            return dec_out  # [B, L, D]
        if self.task_name == 'anomaly_detection':
            dec_out = self.anomaly_detection(x_enc)
            return dec_out  # [B, L, D]
        if self.task_name == 'classification':
            dec_out = self.classification(x_enc, x_mark_enc)
            return dec_out  # [B, N]
        return None