196 lines
6.7 KiB
Python
Executable File
196 lines
6.7 KiB
Python
Executable File
import torch
|
||
import torch.nn as nn
|
||
import torch.nn.functional as F
|
||
|
||
"""
|
||
在原 EXP 模型基础上，添加 Haar 小波变换实现的一层小波去噪，增强时序特征。
|
||
"""
|
||
|
||
class WaveletDenoise(nn.Module):
|
||
"""
|
||
单层 Haar 小波去噪：
|
||
- 使用低通滤波器提取近似系数
|
||
- 通过转置卷积重构时序信号
|
||
"""
|
||
def __init__(self):
|
||
super().__init__()
|
||
# Haar 低通滤波器 [1/√2, 1/√2]
|
||
lp = torch.tensor([1.0, 1.0]) / (2**0.5)
|
||
self.register_buffer('lp_filter', lp.view(1, 1, 2))
|
||
# 转置卷积同滤波器
|
||
self.register_buffer('lp_rec', lp.view(1, 1, 2))
|
||
|
||
def forward(self, x):
|
||
"""
|
||
x: (B, T, N)
|
||
返回去噪后的 (B, T, N)
|
||
"""
|
||
B, T, N = x.shape
|
||
# reshape for conv1d: (B*N, 1, T)
|
||
x_flat = x.permute(0,2,1).contiguous().view(-1, 1, T)
|
||
# 分解
|
||
cA = F.conv1d(x_flat, self.lp_filter, stride=2, padding=0)
|
||
# 重构
|
||
# 反卷积: stride=2, output_padding=T%2
|
||
out = F.conv_transpose1d(cA, self.lp_rec, stride=2,
|
||
output_padding=(T % 2))
|
||
# 裁剪至原始长度
|
||
out = out[:, :, :T]
|
||
# reshape back
|
||
x_dn = out.view(B, N, T).permute(0,2,1)
|
||
return x_dn
|
||
|
||
class DynamicGraphConstructor(nn.Module):
|
||
def __init__(self, node_num, embed_dim):
|
||
super().__init__()
|
||
self.nodevec1 = nn.Parameter(torch.randn(node_num, embed_dim), requires_grad=True)
|
||
self.nodevec2 = nn.Parameter(torch.randn(node_num, embed_dim), requires_grad=True)
|
||
|
||
def forward(self):
|
||
adj = torch.matmul(self.nodevec1, self.nodevec2.T)
|
||
adj = F.relu(adj)
|
||
adj = F.softmax(adj, dim=-1)
|
||
return adj
|
||
|
||
class GraphConvBlock(nn.Module):
|
||
def __init__(self, input_dim, output_dim):
|
||
super().__init__()
|
||
self.theta = nn.Linear(input_dim, output_dim)
|
||
self.residual = (input_dim == output_dim)
|
||
if not self.residual:
|
||
self.res_proj = nn.Linear(input_dim, output_dim)
|
||
|
||
def forward(self, x, adj):
|
||
res = x
|
||
x = torch.matmul(adj, x)
|
||
x = self.theta(x)
|
||
x = x + (res if self.residual else self.res_proj(res))
|
||
return F.relu(x)
|
||
|
||
class MANBA_Block(nn.Module):
|
||
def __init__(self, input_dim, hidden_dim):
|
||
super().__init__()
|
||
self.attn = nn.MultiheadAttention(embed_dim=input_dim, num_heads=4, batch_first=True)
|
||
self.ffn = nn.Sequential(
|
||
nn.Linear(input_dim, hidden_dim),
|
||
nn.ReLU(),
|
||
nn.Linear(hidden_dim, input_dim)
|
||
)
|
||
self.norm1 = nn.LayerNorm(input_dim)
|
||
self.norm2 = nn.LayerNorm(input_dim)
|
||
|
||
def forward(self, x):
|
||
res = x
|
||
x_attn, _ = self.attn(x, x, x)
|
||
x = self.norm1(res + x_attn)
|
||
res2 = x
|
||
x_ffn = self.ffn(x)
|
||
x = self.norm2(res2 + x_ffn)
|
||
return x
|
||
|
||
class SandwichBlock(nn.Module):
|
||
def __init__(self, num_nodes, embed_dim, hidden_dim):
|
||
super().__init__()
|
||
self.manba1 = MANBA_Block(hidden_dim, hidden_dim * 2)
|
||
self.graph_constructor = DynamicGraphConstructor(num_nodes, embed_dim)
|
||
self.gc = GraphConvBlock(hidden_dim, hidden_dim)
|
||
self.manba2 = MANBA_Block(hidden_dim, hidden_dim * 2)
|
||
|
||
def forward(self, h):
|
||
h1 = self.manba1(h)
|
||
adj = self.graph_constructor()
|
||
h2 = self.gc(h1, adj)
|
||
h3 = self.manba2(h2)
|
||
return h3
|
||
|
||
class MLP(nn.Module):
|
||
def __init__(self, in_dim, hidden_dims, out_dim, activation=nn.ReLU):
|
||
super().__init__()
|
||
dims = [in_dim] + hidden_dims + [out_dim]
|
||
layers = []
|
||
for i in range(len(dims)-2):
|
||
layers += [nn.Linear(dims[i], dims[i+1]), activation()]
|
||
layers += [nn.Linear(dims[-2], dims[-1])]
|
||
self.net = nn.Sequential(*layers)
|
||
|
||
def forward(self, x):
|
||
return self.net(x)
|
||
|
||
class EXP(nn.Module):
|
||
def __init__(self, args):
|
||
super().__init__()
|
||
self.horizon = args['horizon']
|
||
self.output_dim = args['output_dim']
|
||
self.seq_len = args.get('in_len', 12)
|
||
self.hidden_dim = args.get('hidden_dim', 64)
|
||
self.num_nodes = args['num_nodes']
|
||
self.embed_dim = args.get('embed_dim', 16)
|
||
|
||
# ==== NEW: discrete time embeddings ====
|
||
self.time_slots = args.get('time_slots', 24 * 60 // args.get('time_slot', 5))
|
||
self.time_embedding = nn.Embedding(self.time_slots, self.hidden_dim)
|
||
self.day_embedding = nn.Embedding(7, self.hidden_dim)
|
||
|
||
# ==== NEW: 小波去噪层 ====
|
||
self.wavelet = WaveletDenoise()
|
||
|
||
# input projection now only takes the denoised flow history
|
||
self.input_proj = MLP(
|
||
in_dim = self.seq_len,
|
||
hidden_dims = [self.hidden_dim],
|
||
out_dim = self.hidden_dim
|
||
)
|
||
|
||
# two Sandwich blocks remain unchanged
|
||
self.sandwich1 = SandwichBlock(self.num_nodes, self.embed_dim, self.hidden_dim)
|
||
self.sandwich2 = SandwichBlock(self.num_nodes, self.embed_dim, self.hidden_dim)
|
||
|
||
# output projection unchanged
|
||
self.out_proj = MLP(
|
||
in_dim = self.hidden_dim,
|
||
hidden_dims = [2 * self.hidden_dim],
|
||
out_dim = self.horizon * self.output_dim
|
||
)
|
||
|
||
def forward(self, x):
|
||
"""
|
||
x: (B, T, N, D_total)
|
||
D_total >= 3 where
|
||
x[...,0] = flow,
|
||
x[...,1] = time_in_day (0 … 1),
|
||
x[...,2] = day_in_week (0–6)
|
||
"""
|
||
x_flow = x[..., 0] # (B, T, N)
|
||
x_time = x[..., 1] # (B, T, N)
|
||
x_day = x[..., 2] # (B, T, N)
|
||
|
||
B, T, N = x_flow.shape
|
||
assert T == self.seq_len
|
||
|
||
# 1) 小波去噪
|
||
x_dn = self.wavelet(x_flow) # (B, T, N)
|
||
|
||
# 2) project the denoised flow history
|
||
x_flat = x_dn.permute(0, 2, 1).reshape(B * N, T)
|
||
h0 = self.input_proj(x_flat).view(B, N, self.hidden_dim)
|
||
|
||
# 3) lookup discrete time indexes at the last time step
|
||
t_idx = (x_time[:, -1, :,] * (self.time_slots - 1)).long()
|
||
d_idx = x_day[:, -1, :,].long()
|
||
time_emb = self.time_embedding(t_idx) # (B, N, hidden_dim)
|
||
day_emb = self.day_embedding(d_idx) # (B, N, hidden_dim)
|
||
|
||
# 4) inject them into the initial hidden state
|
||
h0 = h0 + time_emb + day_emb
|
||
|
||
# 5) the usual Sandwich + residuals
|
||
h1 = self.sandwich1(h0)
|
||
h1 = h1 + h0
|
||
h2 = self.sandwich2(h1)
|
||
|
||
# 6) output projection
|
||
out = self.out_proj(h2) # (B, N, horizon*output_dim)
|
||
out = out.view(B, N, self.horizon, self.output_dim)
|
||
out = out.permute(0, 2, 1, 3) # (B, horizon, N, output_dim)
|
||
return out
|