使用DeeperGCN训练和测试SEED数据集

原创

已于 2023-04-09 14:16:09 修改 · 666 阅读

1 ·

CC 4.0 BY-SA版权

文章标签：

#深度学习 #pytorch #python

于 2023-02-15 22:00:08 首次发布

文章介绍了一个用PyTorch实现的ECG_DeeperGCNs模型，该模型基于图神经网络（GENConv）进行心电图数据的分析。模型包含多层DeepGCNLayer，用于学习节点特征，并使用Adam优化器和交叉熵损失函数进行训练。代码还定义了训练和测试过程，以及数据加载器。模型在SEED数据集上进行训练和评估。

# -*- coding: utf-8 -*-
#
# Copyright (C) 2022 Emperor_Yang, Inc. All Rights Reserved 
#
# @CreateTime    : 2023/2/9 22:12
# @Author        : Emperor_Yang 
# @File          : ECG_DeepGCNs.py
# @Software      : PyCharm

import torch
from torch.nn import ReLU
from easydict import EasyDict
from torch_geometric.nn import GENConv, LayerNorm, DeepGCNLayer, global_add_pool
from torch_geometric.data import DataLoader
from data_process.seed_loader_gnn_memory import SeedGnnMemoryDataset

config = EasyDict()
config.learn_rate = 0.01
config.epoch = 5
config.note_feature_dim = 5
config.note_num = 62
config.hidden_channels = 16
config.class_num = 3
config.hidden_layers = 2
config.batch_size = 16
config.max_loss_increase_time = 3


class ECG_DeeperGCNs(torch.nn.Module):
    """
    GCN handle ECG
    """

    def __init__(self, in_channels, hidden_channels, out_channels):
        super(ECG_DeeperGCNs, self).__init__(