时间序列（Time-Series）TimesNet.py代码解析-Toy模板网

这篇具有很好参考价值的文章主要介绍了时间序列（Time-Series）TimesNet.py代码解析。希望对大家有所帮助。如果存在错误或未考虑完全的地方，请大家不吝赐教，您也可以点击"举报违法"按钮提交疑问。

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.fft
from layers.Embed import DataEmbedding
from layers.Conv_Blocks import Inception_Block_V1

#定义一个用于执行傅里叶变换的函数，并根据傅里叶变换的振幅找到数据的周期
def FFT_for_Period(x, k=2):
# [B, T, C]
xf = torch.fft.rfft(x, dim=1) #执行实数的快速傅里叶变换
# find period by amplitudes

frequency_list = abs(xf).mean(0).mean(-1) #计算变换后振幅的均值

frequency_list[0] = 0 #忽略直流分量
_, top_list = torch.topk(frequency_list, k) #找到top k个频率成分
top_list = top_list.detach().cpu().numpy() #从gpu转移cpu并转换为numpy数组
period = x.shape[1] // top_list #计算周期
return period, abs(xf).mean(-1)[:, top_list] #返回周期和这些周期对应的平均振幅

#定义神经网络模块，继承nn父类，用于处理时间序列中的时间周期变化
class TimesBlock(nn.Module):
def __init__(self, configs):
#调用父类初始化函数
super(TimesBlock, self).__init__()
#配置参数
self.seq_len = configs.seq_len #序列长度
self.pred_len = configs.pred_len #预测长度
self.k = configs.top_k #顶部元素数量
# parameter-efficient design
#构建一个卷积网络作为模型的一部分
self.conv = nn.Sequential(
Inception_Block_V1(configs.d_model, configs.d_ff,
num_kernels=configs.num_kernels),
nn.GELU(),
Inception_Block_V1(configs.d_ff, configs.d_model,
num_kernels=configs.num_kernels)
)
#前向传播函数，处理时间序列x，并返回经过处理的序列
def forward(self, x):
B, T, N = x.size()
period_list, period_weight = FFT_for_Period(x, self.k)

res = []
for i in range(self.k):
period = period_list[i]
# padding
if (self.seq_len + self.pred_len) % period != 0:
length = (
((self.seq_len + self.pred_len) // period) + 1) * period
padding = torch.zeros([x.shape[0], (length - (self.seq_len + self.pred_len)), x.shape[2]]).to(x.device)
out = torch.cat([x, padding], dim=1)
else:
length = (self.seq_len + self.pred_len)
out = x
# reshape
out = out.reshape(B, length // period, period,
N).permute(0, 3, 1, 2).contiguous()
# 2D conv: from 1d Variation to 2d Variation
out = self.conv(out)
# reshape back
out = out.permute(0, 2, 3, 1).reshape(B, -1, N)
res.append(out[:, :(self.seq_len + self.pred_len), :])
res = torch.stack(res, dim=-1)
# adaptive aggregation
period_weight = F.softmax(period_weight, dim=1)
period_weight = period_weight.unsqueeze(
1).unsqueeze(1).repeat(1, T, N, 1)
res = torch.sum(res * period_weight, -1)
# residual connection
res = res + x
return res

#model类
class Model(nn.Module):
"""
Paper link: https://openreview.net/pdf?id=ju_Uqw384Oq
"""
#构造函数，初始化数据
def __init__(self, configs):
super(Model, self).__init__()
self.configs = configs
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.model = nn.ModuleList([TimesBlock(configs)
for _ in range(configs.e_layers)])
self.enc_embedding = DataEmbedding(configs.enc_in, configs.d_model, configs.embed, configs.freq,
configs.dropout)
self.layer = configs.e_layers
self.layer_norm = nn.LayerNorm(configs.d_model)
if self.task_name == 'long_term_forecast' or self.task_name == 'short_term_forecast':
self.predict_linear = nn.Linear(
self.seq_len, self.pred_len + self.seq_len)
self.projection = nn.Linear(
configs.d_model, configs.c_out, bias=True)
if self.task_name == 'imputation' or 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)
#预测函数，用于长短期预测任务
def forecast(self, x_enc, x_mark_enc, x_dec, x_mark_dec):
# Normalization from Non-stationary Transformer
means = x_enc.mean(1, keepdim=True).detach()
x_enc = x_enc - means
stdev = torch.sqrt(
torch.var(x_enc, dim=1, keepdim=True, unbiased=False) + 1e-5)
x_enc /= stdev

# embedding
enc_out = self.enc_embedding(x_enc, x_mark_enc) # [B,T,C]
enc_out = self.predict_linear(enc_out.permute(0, 2, 1)).permute(
0, 2, 1) # align temporal dimension
# TimesNet
for i in range(self.layer):
enc_out = self.layer_norm(self.model[i](enc_out))
# porject back
dec_out = self.projection(enc_out)

# De-Normalization from Non-stationary Transformer
dec_out = dec_out * \
(stdev[:, 0, :].unsqueeze(1).repeat(
1, self.pred_len + self.seq_len, 1))
dec_out = dec_out + \
(means[:, 0, :].unsqueeze(1).repeat(
1, self.pred_len + self.seq_len, 1))
return dec_out
#插补函数，用于数据插补
def imputation(self, x_enc, x_mark_enc, x_dec, x_mark_dec, mask):
# Normalization from Non-stationary Transformer
means = torch.sum(x_enc, dim=1) / torch.sum(mask == 1, dim=1)
means = means.unsqueeze(1).detach()
x_enc = x_enc - means
x_enc = x_enc.masked_fill(mask == 0, 0)
stdev = torch.sqrt(torch.sum(x_enc * x_enc, dim=1) /
torch.sum(mask == 1, dim=1) + 1e-5)
stdev = stdev.unsqueeze(1).detach()
x_enc /= stdev

# embedding
enc_out = self.enc_embedding(x_enc, x_mark_enc) # [B,T,C]
# TimesNet
for i in range(self.layer):
enc_out = self.layer_norm(self.model[i](enc_out))
# porject back
dec_out = self.projection(enc_out)

# De-Normalization from Non-stationary Transformer
dec_out = dec_out * \
(stdev[:, 0, :].unsqueeze(1).repeat(
1, self.pred_len + self.seq_len, 1))
dec_out = dec_out + \
(means[:, 0, :].unsqueeze(1).repeat(
1, self.pred_len + self.seq_len, 1))
return dec_out
#异常检测函数，用于异常检测任务
def anomaly_detection(self, x_enc):
# Normalization from Non-stationary Transformer
means = x_enc.mean(1, keepdim=True).detach()
x_enc = x_enc - means
stdev = torch.sqrt(
torch.var(x_enc, dim=1, keepdim=True, unbiased=False) + 1e-5)
x_enc /= stdev

# embedding
enc_out = self.enc_embedding(x_enc, None) # [B,T,C]
# TimesNet
for i in range(self.layer):
enc_out = self.layer_norm(self.model[i](enc_out))
# porject back
dec_out = self.projection(enc_out)

# Output
# the output transformer encoder/decoder embeddings don't include non-linearity
output = self.act(enc_out)
output = self.dropout(output)
# zero-out padding embeddings
output = output * x_mark_enc.unsqueeze(-1)
# (batch_size, seq_length * d_model)
output = output.reshape(output.shape[0], -1)
output = self.projection(output) # (batch_size, num_classes)
return output
#根据任务类型选择相应的处理函数
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
文章来源地址https://www.toymoban.com/news/detail-820942.html