更新v2可读性
This commit is contained in:
parent
0ad1494aec
commit
f3b5bdfc28
|
|
@ -1,7 +1,7 @@
|
||||||
basic:
|
basic:
|
||||||
dataset: "SolarEnergy"
|
dataset: "SolarEnergy"
|
||||||
mode : "train"
|
mode : "train"
|
||||||
device : "cuda:1"
|
device : "cuda:0"
|
||||||
model: "AEPSA_v2"
|
model: "AEPSA_v2"
|
||||||
seed: 2023
|
seed: 2023
|
||||||
|
|
||||||
|
|
|
||||||
|
|
@ -6,311 +6,150 @@ from model.AEPSA.normalizer import GumbelSoftmax
|
||||||
from model.AEPSA.reprogramming import ReprogrammingLayer
|
from model.AEPSA.reprogramming import ReprogrammingLayer
|
||||||
import torch.nn.functional as F
|
import torch.nn.functional as F
|
||||||
|
|
||||||
# 该文件实现了基于动态图增强的时空序列预测模型
|
# 基于动态图增强的时空序列预测模型实现
|
||||||
# 主要包含三个类：DynamicGraphEnhancer（动态图增强器）、GraphEnhancedEncoder（图增强编码器）和AEPSA（主模型）
|
|
||||||
# 每个操作都标注了输入输出shape以帮助理解数据流向
|
|
||||||
|
|
||||||
class DynamicGraphEnhancer(nn.Module):
|
class DynamicGraphEnhancer(nn.Module):
|
||||||
"""
|
"""动态图增强编码器"""
|
||||||
动态图增强器，基于节点嵌入自动生成图结构
|
|
||||||
参考DDGCRN的设计，使用节点嵌入和特征信息动态计算邻接矩阵
|
|
||||||
"""
|
|
||||||
def __init__(self, num_nodes, in_dim, embed_dim=10):
|
def __init__(self, num_nodes, in_dim, embed_dim=10):
|
||||||
# num_nodes: 节点数量
|
|
||||||
# in_dim: 输入特征维度
|
|
||||||
# embed_dim: 节点嵌入维度
|
|
||||||
super().__init__()
|
super().__init__()
|
||||||
self.num_nodes = num_nodes
|
self.num_nodes = num_nodes # 节点个数
|
||||||
self.embed_dim = embed_dim
|
self.embed_dim = embed_dim # 节点嵌入维度
|
||||||
|
|
||||||
# 节点嵌入参数 [num_nodes, embed_dim]
|
self.node_embeddings = nn.Parameter(torch.randn(num_nodes, embed_dim), requires_grad=True) # 节点嵌入参数
|
||||||
self.node_embeddings = nn.Parameter(
|
|
||||||
torch.randn(num_nodes, embed_dim), requires_grad=True
|
self.feature_transform = nn.Sequential( # 特征转换网络
|
||||||
|
nn.Linear(in_dim, 16),
|
||||||
|
nn.Sigmoid(),
|
||||||
|
nn.Linear(16, 2),
|
||||||
|
nn.Sigmoid(),
|
||||||
|
nn.Linear(2, embed_dim)
|
||||||
)
|
)
|
||||||
|
|
||||||
# 特征转换层，用于生成动态调整的嵌入
|
self.register_buffer("eye", torch.eye(num_nodes)) # 注册单位矩阵
|
||||||
# 输入: [N, in_dim] -> 输出: [N, embed_dim]
|
|
||||||
self.feature_transform = nn.Sequential(
|
|
||||||
nn.Linear(in_dim, 16), # [N, in_dim] -> [N, 16]
|
|
||||||
nn.Sigmoid(),
|
|
||||||
nn.Linear(16, 2), # [N, 16] -> [N, 2]
|
|
||||||
nn.Sigmoid(),
|
|
||||||
nn.Linear(2, embed_dim) # [N, 2] -> [N, embed_dim]
|
|
||||||
)
|
|
||||||
|
|
||||||
# 注册单位矩阵作为固定的支持矩阵 [num_nodes, num_nodes]
|
|
||||||
self.register_buffer("eye", torch.eye(num_nodes))
|
|
||||||
|
|
||||||
def get_laplacian(self, graph, I, normalize=True):
|
def get_laplacian(self, graph, I, normalize=True):
|
||||||
"""
|
D_inv = torch.diag_embed(torch.sum(graph, -1) ** (-0.5)) # 度矩阵的逆平方根
|
||||||
计算归一化拉普拉斯矩阵
|
|
||||||
|
|
||||||
参数:
|
|
||||||
graph: 邻接矩阵 [N, N]
|
|
||||||
I: 单位矩阵 [N, N]
|
|
||||||
normalize: 是否使用标准化拉普拉斯矩阵
|
|
||||||
|
|
||||||
返回:
|
|
||||||
laplacian: 归一化拉普拉斯矩阵 [N, N]
|
|
||||||
"""
|
|
||||||
# 计算度矩阵的逆平方根 [N, N]
|
|
||||||
D_inv = torch.diag_embed(torch.sum(graph, -1) ** (-0.5)) # [N, N]
|
|
||||||
D_inv[torch.isinf(D_inv)] = 0.0 # 处理零除问题
|
D_inv[torch.isinf(D_inv)] = 0.0 # 处理零除问题
|
||||||
|
|
||||||
if normalize:
|
if normalize:
|
||||||
# 归一化拉普拉斯矩阵: D^(-1/2) * graph * D^(-1/2) [N, N]
|
return torch.matmul(torch.matmul(D_inv, graph), D_inv) # 归一化拉普拉斯矩阵
|
||||||
return torch.matmul(torch.matmul(D_inv, graph), D_inv) # [N, N]
|
|
||||||
else:
|
else:
|
||||||
# 拉普拉斯矩阵: D^(-1/2) * (graph + I) * D^(-1/2) [N, N]
|
return torch.matmul(torch.matmul(D_inv, graph + I), D_inv) # 带自环的归一化拉普拉斯矩阵
|
||||||
return torch.matmul(torch.matmul(D_inv, graph + I), D_inv) # [N, N]
|
|
||||||
|
|
||||||
def forward(self, X):
|
def forward(self, X):
|
||||||
"""
|
"""生成动态拉普拉斯矩阵"""
|
||||||
参数:
|
batch_size = X.size(0) # 批次大小
|
||||||
X: 输入特征 [B, N, D]，其中B为批次大小，N为节点数，D为特征维度
|
laplacians = [] # 存储各批次的拉普拉斯矩阵
|
||||||
|
I = self.eye.to(X.device) # 移动单位矩阵到目标设备
|
||||||
返回:
|
|
||||||
laplacians: 动态生成的归一化拉普拉斯矩阵 [B, N, N]
|
|
||||||
"""
|
|
||||||
batch_size = X.size(0)
|
|
||||||
laplacians = []
|
|
||||||
|
|
||||||
# 获取单位矩阵 [N, N]
|
|
||||||
I = self.eye.to(X.device)
|
|
||||||
|
|
||||||
for b in range(batch_size):
|
for b in range(batch_size):
|
||||||
# 使用特征转换层生成动态嵌入调整因子 [N, embed_dim]
|
filt = self.feature_transform(X[b]) # 特征转换
|
||||||
filt = self.feature_transform(X[b]) # [N, embed_dim]
|
nodevec = torch.tanh(self.node_embeddings * filt) # 计算节点嵌入
|
||||||
|
adj = F.relu(torch.matmul(nodevec, nodevec.transpose(0, 1))) # 计算邻接矩阵
|
||||||
# 计算节点嵌入向量 [N, embed_dim]
|
laplacian = self.get_laplacian(adj, I) # 计算拉普拉斯矩阵
|
||||||
nodevec = torch.tanh(self.node_embeddings * filt) # [N, embed_dim]
|
|
||||||
|
|
||||||
# 通过节点嵌入的点积计算邻接矩阵 [N, N]
|
|
||||||
adj = F.relu(torch.matmul(nodevec, nodevec.transpose(0, 1))) # [N, N]
|
|
||||||
|
|
||||||
# 计算归一化拉普拉斯矩阵 [N, N]
|
|
||||||
laplacian = self.get_laplacian(adj, I) # [N, N]
|
|
||||||
laplacians.append(laplacian)
|
laplacians.append(laplacian)
|
||||||
|
return torch.stack(laplacians, dim=0) # 堆叠并返回
|
||||||
# 堆叠所有批次的拉普拉斯矩阵 [B, N, N]
|
|
||||||
return torch.stack(laplacians, dim=0) # [B, N, N]
|
|
||||||
|
|
||||||
class GraphEnhancedEncoder(nn.Module):
|
class GraphEnhancedEncoder(nn.Module):
|
||||||
"""
|
"""图增强编码器"""
|
||||||
基于Chebyshev多项式和动态拉普拉斯矩阵的图增强编码器
|
|
||||||
用于将动态图结构信息整合到特征编码中
|
|
||||||
优化版本：支持直接处理原始时间序列输入
|
|
||||||
"""
|
|
||||||
def __init__(self, K=3, in_dim=64, hidden_dim=32, num_nodes=325, embed_dim=10, device='cpu',
|
def __init__(self, K=3, in_dim=64, hidden_dim=32, num_nodes=325, embed_dim=10, device='cpu',
|
||||||
temporal_dim=12, num_features=1):
|
temporal_dim=12, num_features=1):
|
||||||
# K: Chebyshev多项式阶数
|
|
||||||
# in_dim: 输入特征维度
|
|
||||||
# hidden_dim: 隐藏层维度
|
|
||||||
# num_nodes: 节点数量
|
|
||||||
# embed_dim: 嵌入维度
|
|
||||||
# temporal_dim: 时间序列长度
|
|
||||||
# num_features: 特征通道数量
|
|
||||||
super().__init__()
|
super().__init__()
|
||||||
self.K = K # Chebyshev多项式阶数
|
self.K = K # Chebyshev多项式阶数
|
||||||
self.in_dim = in_dim
|
self.in_dim = in_dim # 输入特征维度
|
||||||
self.hidden_dim = hidden_dim
|
self.hidden_dim = hidden_dim # 隐藏层维度
|
||||||
self.device = device
|
self.device = device # 运行设备
|
||||||
self.temporal_dim = temporal_dim
|
self.temporal_dim = temporal_dim # 时间序列长度
|
||||||
self.num_features = num_features
|
self.num_features = num_features # 特征通道数量
|
||||||
|
|
||||||
# 输入预处理层，适配正确的通道维度
|
self.input_projection = nn.Sequential( # 输入投影层
|
||||||
# 输入: [B, C, N, T] -> 输出: [B, in_dim, N, 1]
|
nn.Conv2d(num_features, 16, kernel_size=(1, 3), padding=(0, 1)),
|
||||||
self.input_projection = nn.Sequential(
|
|
||||||
# 2D卷积：[B, C, N, T] -> [B, 16, N, T]
|
|
||||||
nn.Conv2d(num_features, 16, kernel_size=(1, 3), padding=(0, 1)), # [B, C, N, T] -> [B, 16, N, T]
|
|
||||||
nn.ReLU(),
|
nn.ReLU(),
|
||||||
# 2D卷积：[B, 16, N, T] -> [B, in_dim, N, 1]，时间维度上的全局卷积
|
nn.Conv2d(16, in_dim, kernel_size=(1, temporal_dim)),
|
||||||
nn.Conv2d(16, in_dim, kernel_size=(1, temporal_dim)), # [B, 16, N, T] -> [B, in_dim, N, 1]
|
|
||||||
nn.ReLU()
|
nn.ReLU()
|
||||||
)
|
)
|
||||||
|
|
||||||
# 动态图增强器，用于生成动态拉普拉斯矩阵
|
self.graph_enhancer = DynamicGraphEnhancer(num_nodes, in_dim, embed_dim) # 动态图增强器
|
||||||
# 输入: [B, N, in_dim] -> 输出: [B, N, N]
|
self.alpha = nn.Parameter(torch.randn(K + 1, 1)) # 谱系数
|
||||||
self.graph_enhancer = DynamicGraphEnhancer(num_nodes, in_dim, embed_dim)
|
self.W = nn.ParameterList([nn.Parameter(torch.randn(in_dim, hidden_dim)) for _ in range(K + 1)]) # 传播权重
|
||||||
|
self.to(device) # 移动到指定设备
|
||||||
# 谱系数 α_k (可学习参数) [K+1, 1]
|
|
||||||
self.alpha = nn.Parameter(torch.randn(K + 1, 1))
|
|
||||||
|
|
||||||
# 传播权重 W_k (可学习参数)
|
|
||||||
# 每个权重将Chebyshev多项式展开的结果从in_dim映射到hidden_dim
|
|
||||||
# 输入: [N, in_dim] -> 输出: [N, hidden_dim]
|
|
||||||
self.W = nn.ParameterList([
|
|
||||||
nn.Parameter(torch.randn(in_dim, hidden_dim)) for _ in range(K + 1)
|
|
||||||
])
|
|
||||||
|
|
||||||
self.to(device)
|
|
||||||
|
|
||||||
def chebyshev_polynomials(self, L_tilde, X):
|
def chebyshev_polynomials(self, L_tilde, X):
|
||||||
"""
|
"""计算Chebyshev多项式展开"""
|
||||||
递归计算Chebyshev多项式展开 [T_0(L_tilde)X, ..., T_K(L_tilde)X]
|
T_k_list = [X] # T_0(X) = X
|
||||||
|
|
||||||
参数:
|
|
||||||
L_tilde: 归一化拉普拉斯矩阵 [N, N]
|
|
||||||
X: 输入特征 [N, in_dim]
|
|
||||||
|
|
||||||
返回:
|
|
||||||
T_k_list: Chebyshev多项式展开列表 [K+1, N, in_dim]
|
|
||||||
"""
|
|
||||||
# T_0(X) = X [N, in_dim]
|
|
||||||
T_k_list = [X]
|
|
||||||
if self.K >= 1:
|
if self.K >= 1:
|
||||||
# T_1(X) = L_tilde * X [N, in_dim]
|
T_k_list.append(torch.matmul(L_tilde, X)) # T_1(X) = L_tilde * X
|
||||||
T_k_list.append(torch.matmul(L_tilde, X))
|
|
||||||
for k in range(2, self.K + 1):
|
for k in range(2, self.K + 1):
|
||||||
# T_k(X) = 2*L_tilde*T_{k-1}(X) - T_{k-2}(X) [N, in_dim]
|
T_k_list.append(2 * torch.matmul(L_tilde, T_k_list[-1]) - T_k_list[-2]) # 递推计算
|
||||||
T_k_list.append(2 * torch.matmul(L_tilde, T_k_list[-1]) - T_k_list[-2])
|
return T_k_list # 返回多项式列表
|
||||||
# 返回Chebyshev多项式展开列表 [K+1, N, in_dim]
|
|
||||||
return T_k_list
|
|
||||||
|
|
||||||
def forward(self, X):
|
def forward(self, X):
|
||||||
"""
|
"""输入特征[B,N,C,T]，返回增强特征[B,N,hidden_dim*(K+1)]"""
|
||||||
参数:
|
batch_size = X.size(0) # 批次大小
|
||||||
X: 输入特征 [B, N, C, T] 或 [B, N, T]（单特征情况）
|
num_nodes = X.size(1) # 节点数量
|
||||||
B: 批次大小, N: 节点数, C: 特征通道数, T: 时间序列长度
|
|
||||||
|
|
||||||
返回:
|
x = X.permute(0, 2, 1, 3) # [B,C,N,T]
|
||||||
增强后的特征 [B, N, hidden_dim*(K+1)]
|
x_proj = self.input_projection(x).squeeze(-1) # [B,in_dim,N]
|
||||||
"""
|
x_proj = x_proj.permute(0, 2, 1) # [B,N,in_dim]
|
||||||
batch_size = X.size(0)
|
|
||||||
num_nodes = X.size(1)
|
|
||||||
|
|
||||||
# 处理不同维度的输入
|
enhanced_features = [] # 存储增强特征
|
||||||
if len(X.shape) == 4: # [B, N, C, T]
|
laplacians = self.graph_enhancer(x_proj) # 生成动态拉普拉斯矩阵
|
||||||
# 输入: [B, N, C, T] -> 输出: [B, C, N, T]
|
|
||||||
# 将输入转换为[B, C, N, T]格式，适合2D卷积
|
|
||||||
x = X.permute(0, 2, 1, 3) # [B, C, N, T]
|
|
||||||
else: # [B, N, T]
|
|
||||||
# 输入: [B, N, T] -> 输出: [B, 1, N, T]
|
|
||||||
# 添加通道维度
|
|
||||||
x = X.unsqueeze(1) # [B, 1, N, T]
|
|
||||||
|
|
||||||
# 使用卷积投影层处理时间维度
|
|
||||||
# 输入: [B, C, N, T] -> 输出: [B, in_dim, N, 1]
|
|
||||||
x_proj = self.input_projection(x)
|
|
||||||
# 输入: [B, in_dim, N, 1] -> 输出: [B, in_dim, N]
|
|
||||||
x_proj = x_proj.squeeze(-1) # [B, in_dim, N]
|
|
||||||
# 输入: [B, in_dim, N] -> 输出: [B, N, in_dim]
|
|
||||||
x_proj = x_proj.permute(0, 2, 1) # [B, N, in_dim]
|
|
||||||
|
|
||||||
enhanced_features = []
|
|
||||||
|
|
||||||
# 动态生成拉普拉斯矩阵
|
|
||||||
# 输入: [B, N, in_dim] -> 输出: [B, N, N]
|
|
||||||
laplacians = self.graph_enhancer(x_proj) # [B, N, N]
|
|
||||||
|
|
||||||
for b in range(batch_size):
|
for b in range(batch_size):
|
||||||
# 获取当前批次的拉普拉斯矩阵 [N, N]
|
L = laplacians[b] # 当前批次的拉普拉斯矩阵
|
||||||
L = laplacians[b] # [N, N]
|
|
||||||
|
|
||||||
# 特征值缩放
|
# 特征值缩放
|
||||||
try:
|
try:
|
||||||
# 计算最大特征值 [1]
|
lambda_max = torch.linalg.eigvalsh(L).max().real # 最大特征值
|
||||||
lambda_max = torch.linalg.eigvalsh(L).max().real # [1]
|
lambda_max = 1.0 if lambda_max < 1e-6 else lambda_max # 防止除零
|
||||||
# 避免除零问题
|
L_tilde = (2.0 / lambda_max) * L - torch.eye(L.size(0), device=L.device) # 归一化拉普拉斯
|
||||||
if lambda_max < 1e-6:
|
|
||||||
lambda_max = 1.0
|
|
||||||
# 缩放拉普拉斯矩阵到[-1, 1]区间 [N, N]
|
|
||||||
L_tilde = (2.0 / lambda_max) * L - torch.eye(L.size(0), device=L.device) # [N, N]
|
|
||||||
except:
|
except:
|
||||||
# 如果计算特征值失败，使用单位矩阵 [N, N]
|
L_tilde = torch.eye(num_nodes, device=X.device) # 异常处理
|
||||||
L_tilde = torch.eye(num_nodes, device=X.device) # [N, N]
|
|
||||||
|
|
||||||
# 计算Chebyshev多项式展开
|
# 计算展开并应用权重
|
||||||
# 输入: L_tilde [N, N], x_proj [N, in_dim] -> 输出: [K+1, N, in_dim]
|
T_k_list = self.chebyshev_polynomials(L_tilde, x_proj[b]) # 计算Chebyshev多项式
|
||||||
T_k_list = self.chebyshev_polynomials(L_tilde, x_proj[b]) # [K+1, N, in_dim]
|
H_list = [torch.matmul(T_k_list[k], self.W[k]) for k in range(self.K + 1)] # 应用权重
|
||||||
H_list = []
|
X_enhanced = torch.cat(H_list, dim=-1) # 拼接特征
|
||||||
|
|
||||||
# 应用传播权重
|
|
||||||
for k in range(self.K + 1):
|
|
||||||
# 矩阵乘法: [N, in_dim] × [in_dim, hidden_dim] -> [N, hidden_dim]
|
|
||||||
H_k = torch.matmul(T_k_list[k], self.W[k]) # [N, hidden_dim]
|
|
||||||
H_list.append(H_k)
|
|
||||||
|
|
||||||
# 拼接所有尺度的特征
|
|
||||||
# 输入: [K+1, N, hidden_dim] -> 输出: [N, hidden_dim*(K+1)]
|
|
||||||
X_enhanced = torch.cat(H_list, dim=-1) # [N, hidden_dim*(K+1)]
|
|
||||||
enhanced_features.append(X_enhanced)
|
enhanced_features.append(X_enhanced)
|
||||||
|
|
||||||
# 堆叠所有批次的增强特征
|
return torch.stack(enhanced_features, dim=0) # 堆叠返回[B,N,hidden_dim*(K+1)]，每个节点在每个k阶下的切比雪夫特征
|
||||||
# 输入: [B, N, hidden_dim*(K+1)] -> 输出: [B, N, hidden_dim*(K+1)]
|
|
||||||
return torch.stack(enhanced_features, dim=0) # [B, N, hidden_dim*(K+1)]
|
|
||||||
|
|
||||||
class AEPSA(nn.Module):
|
class AEPSA(nn.Module):
|
||||||
"""
|
"""自适应特征投影时空自注意力模型"""
|
||||||
自适应特征投影时空自注意力模型（AEPSA）
|
|
||||||
整合动态图增强和预训练语言模型进行时空序列预测
|
|
||||||
"""
|
|
||||||
def __init__(self, configs):
|
def __init__(self, configs):
|
||||||
# configs: 包含模型所有配置的字典
|
|
||||||
# 主要配置参数说明：
|
|
||||||
# device: 运行设备
|
|
||||||
# pred_len: 预测序列长度
|
|
||||||
# seq_len: 输入序列长度
|
|
||||||
# patch_len: 补丁长度（已移除对应组件）
|
|
||||||
# input_dim: 输入特征维度
|
|
||||||
# stride: 步长（已移除对应组件）
|
|
||||||
# dropout: Dropout概率
|
|
||||||
# gpt_layers: 使用的GPT2层数
|
|
||||||
# d_ff: 前馈网络隐藏层维度
|
|
||||||
# gpt_path: 预训练GPT2模型路径
|
|
||||||
# num_nodes: 节点数量
|
|
||||||
# word_num: GumbelSoftmax词汇数量
|
|
||||||
# d_model: 模型维度
|
|
||||||
# n_heads: 注意力头数量
|
|
||||||
# chebyshev_order: Chebyshev多项式阶数
|
|
||||||
# graph_hidden_dim: 图编码器隐藏层维度
|
|
||||||
# graph_embed_dim: 图编码器嵌入维度
|
|
||||||
super(AEPSA, self).__init__()
|
super(AEPSA, self).__init__()
|
||||||
self.device = configs['device']
|
self.device = configs['device'] # 运行设备
|
||||||
self.pred_len = configs['pred_len']
|
self.pred_len = configs['pred_len'] # 预测序列长度
|
||||||
self.seq_len = configs['seq_len']
|
self.seq_len = configs['seq_len'] # 输入序列长度
|
||||||
self.patch_len = configs['patch_len']
|
self.patch_len = configs['patch_len'] # 补丁长度
|
||||||
self.input_dim = configs['input_dim']
|
self.input_dim = configs['input_dim'] # 输入特征维度
|
||||||
self.stride = configs['stride']
|
self.stride = configs['stride'] # 步长
|
||||||
self.dropout = configs['dropout']
|
self.dropout = configs['dropout'] # Dropout概率
|
||||||
self.gpt_layers = configs['gpt_layers']
|
self.gpt_layers = configs['gpt_layers'] # 使用的GPT2层数
|
||||||
self.d_ff = configs['d_ff']
|
self.d_ff = configs['d_ff'] # 前馈网络隐藏层维度
|
||||||
self.gpt_path = configs['gpt_path']
|
self.gpt_path = configs['gpt_path'] # 预训练GPT2模型路径
|
||||||
self.num_nodes = configs.get('num_nodes', 325) # 添加节点数量配置
|
self.num_nodes = configs.get('num_nodes', 325) # 节点数量
|
||||||
|
|
||||||
# GumbelSoftmax层，用于词汇选择
|
self.word_choice = GumbelSoftmax(configs['word_num']) # 词汇选择层
|
||||||
# 输入: [vocab_size] -> 输出: [vocab_size]（one-hot近似分布）
|
|
||||||
self.word_choice = GumbelSoftmax(configs['word_num'])
|
|
||||||
|
|
||||||
self.d_model = configs['d_model']
|
self.d_model = configs['d_model'] # 模型维度
|
||||||
self.n_heads = configs['n_heads']
|
self.n_heads = configs['n_heads'] # 注意力头数量
|
||||||
self.d_keys = None
|
self.d_keys = None # 键维度
|
||||||
self.d_llm = 768 # GPT2隐藏层维度
|
self.d_llm = 768 # GPT2隐藏层维度
|
||||||
|
|
||||||
self.patch_nums = int((self.seq_len - self.patch_len) / self.stride + 2)
|
self.patch_nums = int((self.seq_len - self.patch_len) / self.stride + 2) # 补丁数量
|
||||||
self.head_nf = self.d_ff * self.patch_nums
|
self.head_nf = self.d_ff * self.patch_nums # 头特征维度
|
||||||
|
|
||||||
# 移除不再使用的patch_embedding层
|
# 初始化GPT2模型
|
||||||
|
self.gpts = GPT2Model.from_pretrained(self.gpt_path, output_attentions=True, output_hidden_states=True) # GPT2模型
|
||||||
# GPT2初始化
|
|
||||||
# 加载预训练GPT2模型，输出注意力权重和隐藏状态
|
|
||||||
self.gpts = GPT2Model.from_pretrained(self.gpt_path, output_attentions=True, output_hidden_states=True)
|
|
||||||
self.gpts.h = self.gpts.h[:self.gpt_layers] # 截取指定层数
|
self.gpts.h = self.gpts.h[:self.gpt_layers] # 截取指定层数
|
||||||
self.gpts.apply(self.reset_parameters)
|
self.gpts.apply(self.reset_parameters) # 重置参数
|
||||||
|
|
||||||
# 获取GPT2词嵌入权重
|
self.word_embeddings = self.gpts.get_input_embeddings().weight.to(self.device) # 词嵌入权重
|
||||||
# 形状: [vocab_size, d_llm]
|
self.vocab_size = self.word_embeddings.shape[0] # 词汇表大小
|
||||||
self.word_embeddings = self.gpts.get_input_embeddings().weight.to(self.device)
|
self.mapping_layer = nn.Linear(self.vocab_size, 1) # 映射层
|
||||||
self.vocab_size = self.word_embeddings.shape[0]
|
self.reprogramming_layer = ReprogrammingLayer(self.d_model, self.n_heads, self.d_keys, self.d_llm) # 重编程层
|
||||||
# 映射层，将词汇表维度映射到1维
|
|
||||||
# 输入: [vocab_size] -> 输出: [1]
|
|
||||||
self.mapping_layer = nn.Linear(self.vocab_size, 1)
|
|
||||||
# 重编程层，用于特征映射和注意力计算
|
|
||||||
# 输入: [B, N, d_model], [d_llm], [d_llm] -> 输出: [B, N, d_model]
|
|
||||||
self.reprogramming_layer = ReprogrammingLayer(self.d_model, self.n_heads, self.d_keys, self.d_llm)
|
|
||||||
|
|
||||||
# 动态图增强编码器
|
# 初始化图增强编码器
|
||||||
# 输入: [B, N, C, T] -> 输出: [B, N, hidden_dim*(K+1)]
|
|
||||||
self.graph_encoder = GraphEnhancedEncoder(
|
self.graph_encoder = GraphEnhancedEncoder(
|
||||||
K=configs.get('chebyshev_order', 3), # Chebyshev多项式阶数
|
K=configs.get('chebyshev_order', 3), # Chebyshev多项式阶数
|
||||||
in_dim=self.d_model, # 输入特征维度
|
in_dim=self.d_model, # 输入特征维度
|
||||||
|
|
@ -322,11 +161,9 @@ class AEPSA(nn.Module):
|
||||||
num_features=self.input_dim # 特征通道数
|
num_features=self.input_dim # 特征通道数
|
||||||
)
|
)
|
||||||
|
|
||||||
# 图特征投影层，将图增强特征维度转换为d_model
|
self.graph_projection = nn.Linear( # 图特征投影层，每一k阶的切比雪夫权重映射到隐藏维度
|
||||||
# 输入: [B, N, hidden_dim*(K+1)] -> 输出: [B, N, d_model]
|
configs.get('graph_hidden_dim', 32) * (configs.get('chebyshev_order', 3) + 1), # 输入维度
|
||||||
self.graph_projection = nn.Linear(
|
self.d_model # 输出维度
|
||||||
configs.get('graph_hidden_dim', 32) * (configs.get('chebyshev_order', 3) + 1),
|
|
||||||
self.d_model
|
|
||||||
)
|
)
|
||||||
|
|
||||||
self.out_mlp = nn.Sequential(
|
self.out_mlp = nn.Sequential(
|
||||||
|
|
@ -335,11 +172,9 @@ class AEPSA(nn.Module):
|
||||||
nn.Linear(128, self.pred_len)
|
nn.Linear(128, self.pred_len)
|
||||||
)
|
)
|
||||||
|
|
||||||
for i, (name, param) in enumerate(self.gpts.named_parameters()):
|
# 设置参数可训练性 wps=word position embeddings
|
||||||
if 'wpe' in name:
|
for name, param in self.gpts.named_parameters():
|
||||||
param.requires_grad = True
|
param.requires_grad = 'wpe' in name
|
||||||
else:
|
|
||||||
param.requires_grad = False
|
|
||||||
|
|
||||||
def reset_parameters(self, module):
|
def reset_parameters(self, module):
|
||||||
if hasattr(module, 'weight') and module.weight is not None:
|
if hasattr(module, 'weight') and module.weight is not None:
|
||||||
|
|
@ -348,60 +183,26 @@ class AEPSA(nn.Module):
|
||||||
torch.nn.init.zeros_(module.bias)
|
torch.nn.init.zeros_(module.bias)
|
||||||
|
|
||||||
def forward(self, x):
|
def forward(self, x):
|
||||||
"""
|
# 数据处理
|
||||||
前向传播函数
|
x = x[..., :1] # [B,T,N,1]
|
||||||
输入:
|
x_enc = rearrange(x, 'b t n c -> b n c t') # [B,N,1,T]
|
||||||
x: 输入数据 [B, T, N, C]，其中B为批次大小，T为时间步长，N为节点数，C为特征通道数
|
|
||||||
|
|
||||||
返回:
|
# 图编码
|
||||||
outputs: 预测结果 [B, pred_len, N, 1]
|
graph_enhanced = self.graph_encoder(x_enc) # [B,N,1,T] -> [B, N, hidden_dim*(K+1)]
|
||||||
"""
|
enc_out = self.graph_projection(graph_enhanced) # [B,N,d_model]
|
||||||
# 只保留第一个特征通道
|
|
||||||
# 输入: [B, T, N, C] -> 输出: [B, T, N, 1]
|
|
||||||
x = x[..., :1] # [B, T, N, 1]
|
|
||||||
|
|
||||||
# 调整输入维度以适配图编码器
|
# 词嵌入处理
|
||||||
# 输入: [B, T, N, 1] -> 输出: [B, N, 1, T]
|
self.mapping_layer(self.word_embeddings.permute(1, 0)).permute(1, 0)
|
||||||
x_enc = rearrange(x, 'b t n c -> b n c t') # [B, N, 1, T]
|
masks = self.word_choice(self.mapping_layer.weight.data.permute(1,0)) # [d_llm,1]
|
||||||
|
source_embeddings = self.word_embeddings[masks==1] # [selected_words,d_llm]
|
||||||
|
|
||||||
# 应用图增强编码器获取增强特征
|
# 重编程与预测
|
||||||
# 输入: [B, N, 1, T] -> 输出: [B, N, hidden_dim*(K+1)]
|
enc_out = self.reprogramming_layer(enc_out, source_embeddings, source_embeddings)
|
||||||
graph_enhanced = self.graph_encoder(x_enc) # [B, N, hidden_dim*(K+1)]
|
enc_out = self.gpts(inputs_embeds=enc_out).last_hidden_state # [B,N,d_llm]
|
||||||
|
dec_out = self.out_mlp(enc_out) # [B,N,pred_len]
|
||||||
|
|
||||||
# 投影图增强特征到模型维度
|
# 维度调整
|
||||||
# 输入: [B, N, hidden_dim*(K+1)] -> 输出: [B, N, d_model]
|
outputs = dec_out.unsqueeze(dim=-1) # [B,N,pred_len,1]
|
||||||
enc_out = self.graph_projection(graph_enhanced) # [B, N, d_model]
|
outputs = outputs.permute(0, 2, 1, 3) # [B,pred_len,N,1]
|
||||||
|
|
||||||
# 处理词嵌入权重，为注意力机制准备
|
return outputs
|
||||||
# 输入: [vocab_size, d_llm] -> 输出: [d_llm, vocab_size] -> [d_llm, vocab_size]
|
|
||||||
self.mapping_layer(self.word_embeddings.permute(1, 0)).permute(1, 0) # [vocab_size, d_llm]
|
|
||||||
|
|
||||||
# 使用GumbelSoftmax选择词汇
|
|
||||||
# 输入: [d_llm, 1] -> 输出: [d_llm, 1]
|
|
||||||
masks = self.word_choice(self.mapping_layer.weight.data.permute(1,0)) # [d_llm, 1]
|
|
||||||
|
|
||||||
# 获取选中的源嵌入
|
|
||||||
# 输入: [vocab_size, d_llm] 与 masks -> 输出: [selected_words, d_llm]
|
|
||||||
source_embeddings = self.word_embeddings[masks==1] # [selected_words, d_llm]
|
|
||||||
|
|
||||||
# 应用重编程层处理特征和源嵌入
|
|
||||||
# 输入: [B, N, d_model], [selected_words, d_llm], [selected_words, d_llm] -> 输出: [B, N, d_model]
|
|
||||||
enc_out = self.reprogramming_layer(enc_out, source_embeddings, source_embeddings) # [B, N, d_model]
|
|
||||||
|
|
||||||
# 通过GPT2模型处理增强特征
|
|
||||||
# 输入: [B, N, d_model] -> 输出: [B, N, d_llm]
|
|
||||||
enc_out = self.gpts(inputs_embeds=enc_out).last_hidden_state # [B, N, d_llm]
|
|
||||||
|
|
||||||
# 使用MLP预测未来时间步
|
|
||||||
# 输入: [B, N, d_llm] -> 输出: [B, N, pred_len]
|
|
||||||
dec_out = self.out_mlp(enc_out) # [B, N, pred_len]
|
|
||||||
|
|
||||||
# 添加通道维度
|
|
||||||
# 输入: [B, N, pred_len] -> 输出: [B, N, pred_len, 1]
|
|
||||||
outputs = dec_out.unsqueeze(dim=-1) # [B, N, pred_len, 1]
|
|
||||||
|
|
||||||
# 调整维度顺序为 [B, pred_len, N, 1]
|
|
||||||
# 输入: [B, N, pred_len, 1] -> 输出: [B, pred_len, N, 1]
|
|
||||||
outputs = outputs.permute(0, 2, 1, 3) # [B, pred_len, N, 1]
|
|
||||||
|
|
||||||
return outputs # [B, pred_len, N, 1]
|
|
||||||
Loading…
Reference in New Issue