File size: 8,587 Bytes
f9d3aeb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import torch
import torch.nn as nn
import torch.nn.functional as F
from layers.Embed import DataEmbedding
from layers.AutoCorrelation import AutoCorrelationLayer
from layers.FourierCorrelation import FourierBlock, FourierCrossAttention
from layers.MultiWaveletCorrelation import MultiWaveletCross, MultiWaveletTransform
from layers.Autoformer_EncDec import Encoder, Decoder, EncoderLayer, DecoderLayer, my_Layernorm, series_decomp


class Model(nn.Module):
    """
    FEDformer performs the attention mechanism on frequency domain and achieved O(N) complexity
    Paper link: https://proceedings.mlr.press/v162/zhou22g.html
    """

    def __init__(self, configs, version='fourier', mode_select='random', modes=32):
        """
        version: str, for FEDformer, there are two versions to choose, options: [Fourier, Wavelets].
        mode_select: str, for FEDformer, there are two mode selection method, options: [random, low].
        modes: int, modes to be selected.
        """
        super(Model, self).__init__()
        self.task_name = configs.task_name
        self.seq_len = configs.seq_len
        self.label_len = configs.label_len
        self.pred_len = configs.pred_len

        self.version = version
        self.mode_select = mode_select
        self.modes = modes

        # Decomp
        self.decomp = series_decomp(configs.moving_avg)
        self.enc_embedding = DataEmbedding(configs.enc_in, configs.d_model, configs.embed, configs.freq,
                                           configs.dropout)
        self.dec_embedding = DataEmbedding(configs.dec_in, configs.d_model, configs.embed, configs.freq,
                                           configs.dropout)

        if self.version == 'Wavelets':
            encoder_self_att = MultiWaveletTransform(ich=configs.d_model, L=1, base='legendre')
            decoder_self_att = MultiWaveletTransform(ich=configs.d_model, L=1, base='legendre')
            decoder_cross_att = MultiWaveletCross(in_channels=configs.d_model,
                                                  out_channels=configs.d_model,
                                                  seq_len_q=self.seq_len // 2 + self.pred_len,
                                                  seq_len_kv=self.seq_len,
                                                  modes=self.modes,
                                                  ich=configs.d_model,
                                                  base='legendre',
                                                  activation='tanh')
        else:
            encoder_self_att = FourierBlock(in_channels=configs.d_model,
                                            out_channels=configs.d_model,
                                            n_heads=configs.n_heads,
                                            seq_len=self.seq_len,
                                            modes=self.modes,
                                            mode_select_method=self.mode_select)
            decoder_self_att = FourierBlock(in_channels=configs.d_model,
                                            out_channels=configs.d_model,
                                            n_heads=configs.n_heads,
                                            seq_len=self.seq_len // 2 + self.pred_len,
                                            modes=self.modes,
                                            mode_select_method=self.mode_select)
            decoder_cross_att = FourierCrossAttention(in_channels=configs.d_model,
                                                      out_channels=configs.d_model,
                                                      seq_len_q=self.seq_len // 2 + self.pred_len,
                                                      seq_len_kv=self.seq_len,
                                                      modes=self.modes,
                                                      mode_select_method=self.mode_select,
                                                      num_heads=configs.n_heads)
        # Encoder
        self.encoder = Encoder(
            [
                EncoderLayer(
                    AutoCorrelationLayer(
                        encoder_self_att,  # instead of multi-head attention in transformer
                        configs.d_model, configs.n_heads),
                    configs.d_model,
                    configs.d_ff,
                    moving_avg=configs.moving_avg,
                    dropout=configs.dropout,
                    activation=configs.activation
                ) for l in range(configs.e_layers)
            ],
            norm_layer=my_Layernorm(configs.d_model)
        )
        # Decoder
        self.decoder = Decoder(
            [
                DecoderLayer(
                    AutoCorrelationLayer(
                        decoder_self_att,
                        configs.d_model, configs.n_heads),
                    AutoCorrelationLayer(
                        decoder_cross_att,
                        configs.d_model, configs.n_heads),
                    configs.d_model,
                    configs.c_out,
                    configs.d_ff,
                    moving_avg=configs.moving_avg,
                    dropout=configs.dropout,
                    activation=configs.activation,
                )
                for l in range(configs.d_layers)
            ],
            norm_layer=my_Layernorm(configs.d_model),
            projection=nn.Linear(configs.d_model, configs.c_out, bias=True)
        )

        if self.task_name == 'imputation':
            self.projection = nn.Linear(configs.d_model, configs.c_out, bias=True)
        if self.task_name == 'anomaly_detection':
            self.projection = nn.Linear(configs.d_model, configs.c_out, bias=True)
        if self.task_name == 'classification':
            self.act = F.gelu
            self.dropout = nn.Dropout(configs.dropout)
            self.projection = nn.Linear(configs.d_model * configs.seq_len, configs.num_class)

    def forecast(self, x_enc, x_mark_enc, x_dec, x_mark_dec):
        # decomp init
        mean = torch.mean(x_enc, dim=1).unsqueeze(1).repeat(1, self.pred_len, 1)
        seasonal_init, trend_init = self.decomp(x_enc)  # x - moving_avg, moving_avg
        # decoder input
        trend_init = torch.cat([trend_init[:, -self.label_len:, :], mean], dim=1)
        seasonal_init = F.pad(seasonal_init[:, -self.label_len:, :], (0, 0, 0, self.pred_len))
        # enc
        enc_out = self.enc_embedding(x_enc, x_mark_enc)
        dec_out = self.dec_embedding(seasonal_init, x_mark_dec)
        enc_out, attns = self.encoder(enc_out, attn_mask=None)
        # dec
        seasonal_part, trend_part = self.decoder(dec_out, enc_out, x_mask=None, cross_mask=None, trend=trend_init)
        # final
        dec_out = trend_part + seasonal_part
        return dec_out

    def imputation(self, x_enc, x_mark_enc, x_dec, x_mark_dec, mask):
        # enc
        enc_out = self.enc_embedding(x_enc, x_mark_enc)
        enc_out, attns = self.encoder(enc_out, attn_mask=None)
        # final
        dec_out = self.projection(enc_out)
        return dec_out

    def anomaly_detection(self, x_enc):
        # enc
        enc_out = self.enc_embedding(x_enc, None)
        enc_out, attns = self.encoder(enc_out, attn_mask=None)
        # final
        dec_out = self.projection(enc_out)
        return dec_out

    def classification(self, x_enc, x_mark_enc):
        # enc
        enc_out = self.enc_embedding(x_enc, None)
        enc_out, attns = self.encoder(enc_out, attn_mask=None)

        # Output
        output = self.act(enc_out)
        output = self.dropout(output)
        output = output * x_mark_enc.unsqueeze(-1)
        output = output.reshape(output.shape[0], -1)
        output = self.projection(output)
        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