生成.mat数据：

%创建时间：2025年6月19日
%zhouzhichao
%遍历生成不同程度多径效应影响的无线通信网络拓扑推理数据用于测试close all
clearsnr = 40;
n = 30;dataset_n = 100;for bias = 0.1:0.1:0.9nodes_P = ones(n,1);Sampling_M = 3000;%获取一帧信号及对应功率[ack,~] = ack_generate(Sampling_M);ack_L = length(ack);signal = ack;P_signal = sum(abs(signal).^2);ack_noise = randn(ack_L, 1);P_noise_1 = sum(ack_noise.^2);  % 计算当前噪声的能量a_slot_N = round(7*ack_L);ave_edge = round(1.5*n);L = ave_edge*3*a_slot_N;Signals = zeros(dataset_n,n,L);Tp = zeros(dataset_n,n,n);Tp_list = cell(dataset_n,1);A = 10^(snr/10);P_noise = P_signal/A;alpha = sqrt(P_noise / P_noise_1);  % 计算缩放因子for k=1:dataset_ndisp(["bias: ",bias,"  k: ",k])c = 0;%待生成内容：tp，tp_list，signals[tp,tp_list] = large_nodes_tp(nodes_P,n);signals = zeros(n,L);n_edge = length(tp_list);for epoch=1:3index_list = randperm(n_edge);for i =index_listthis_slot_start_point = c*a_slot_N+1;p = tp_list(i,1);q = tp_list(i,2);P1 = nodes_P(p);P2 = nodes_P(q);%三次响应for m=1:3r_P = 1 - bias + 2*bias * rand;signal = r_P*P1*ack;signals(p,this_slot_start_point+(2*m-2)*ack_L:this_slot_start_point+(2*m-1)*ack_L-1) = signal;r_P = 1 - bias + 2*bias * rand;signal = r_P*P2*ack;signals(q,this_slot_start_point+(2*m-1)*ack_L:this_slot_start_point+(2*m)*ack_L-1) = signal;endc = c + 1;endendfor i=1:nsignals(i,:) = signals(i,:) + alpha*randn(1, L);endSignals(k,:,:) = signals;Tp(k,:,:) = tp;Tp_list{k} = tp_list;endsave("D:\无线通信网络认知\论文1\大修意见\Reviewer2-3 多径传输实验\diff var mat\bias_"+num2str(bias)+"_dataset.mat","Tp","Tp_list","Signals")
end

 

转换为PyG可用数据：

#作者：zhouzhichao
#创建时间：2025/6/19
#内容：生成多径效应影响数据集用于实验import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)
import os
import h5py
import matplotlib.pyplot as plt
import numpy as np
import torch
from torch_geometric.data import InMemoryDataset, Data
from torch_geometric.utils import negative_samplingbase_dir = "D:\无线通信网络认知\论文1\大修意见\Reviewer2-3 多径传输实验\diff var mat\\"N = 30
grapg_size = N
train_n = 20class graph_data(InMemoryDataset):def __init__(self, root, signals=None, tp_list = None, transform=None, pre_transform=None):self.signals = signalsself.tp_list = tp_listsuper().__init__(root, transform, pre_transform)self.data = torch.load(self.processed_paths[0])# 返回process方法所需的保存文件名。你之后保存的数据集名字和列表里的一致@propertydef processed_file_names(self):return ['gcn_data.pt']# 生成数据集所用的方法def process(self):signals = self.signalstp_list =self.tp_list#matlab的index是从1开始的，pytorch的index应当从0开始tp_list = tp_list - 1X = torch.tensor(signals, dtype=torch.float)# 所有的边Edge_index = torch.tensor(tp_list, dtype=torch.long)# 所有的边1标签edge_label = np.ones((tp_list.shape[1]))Edge_label = torch.tensor(edge_label, dtype=torch.float)neg_edge_index = negative_sampling(edge_index=Edge_index, num_nodes=grapg_size,num_neg_samples=Edge_index.shape[1], method='sparse')# 拼接正负样本索引Edge_label_index = Edge_indexperm = torch.randperm(Edge_index.size(1))Edge_index = Edge_index[:, perm]Edge_index = Edge_index[:, :train_n]# neg_edge_index = neg_edge_index[:, :train_n]Edge_label_index = torch.cat([Edge_label_index, neg_edge_index],dim=-1,)# 拼接正负样本Edge_label = torch.cat([Edge_label,Edge_label.new_zeros(neg_edge_index.size(1))], dim=0)data = Data(x=X, edge_index=Edge_index, edge_label_index=Edge_label_index, edge_label=Edge_label)torch.save(data, self.processed_paths[0])# data_list.append(data)# data_, slices = self.collate(data_list)  # 将不同大小的图数据对齐，填充# torch.save((data_, slices), self.processed_paths[0])if __name__ =="__main__":# for b in np.arange(0.1, 1.0, 0.1):#     print("b： ",b)for b in np.arange(0.1, 1.0, 0.1):bias = round(b,1)mat_file = h5py.File(base_dir + 'bias_'+str(bias)+'_dataset.mat', 'r')# 获取数据集Signals = mat_file["Signals"][()]Tp = mat_file["Tp"][()]Tp_list = mat_file["Tp_list"][()]n = 100for i in range(n):signals = Signals[:, :, i]tp_list = np.array(mat_file[Tp_list[0, i]])root = 'diff var graph/bias_'+str(bias)+'_data_'+str(i)graph_data(root, signals = signals, tp_list = tp_list)print("")print("...图数据生成完成...")