在PPI数据集上使用GCN网络实现节点分类

from sklearn.metrics import f1_score import torch_geometric.transforms as T from torch_geometric.datasets import PPI from torch_geometric.loader import DataLoader from models import * import scipy.sparse as sp import numpy as np import matplotlib.pyplot as plt import pandas as pd import pickle import argparse import warnings warnings.filterwarnings("ignore") parser = argparse.ArgumentParser() parser.add_argument('--no-cuda', action='store_true', default=False, help='Disables CUDA training.') parser.add_argument('--fastmode', action='store_true', default=False, help='Validate during training pass.') parser.add_argument('--seed', type=int, default=42, help='Random seed.') parser.add_argument('--epochs', type=int, default=200, help='Number of epochs to train.') parser.add_argument('--lr', type=float, default=0.005, help='Initial learning rate.') parser.add_argument('--weight_decay', type=float, default=5e-4, help='Weight decay (L2 loss on parameters).') parser.add_argument('--hidden', type=int, default=256, help='Number of hidden units.') parser.add_argument('--dropout', type=float, default=0.8, help='Dropout rate (1 - keep probability).') parser.add_argument("--normalization", default="FirstOrderGCN", # FirstOrderGCN help="The normalization on the adj matrix.") parser.add_argument('--dataset', default="ppi", help="The data set") #citeseer cora parser.add_argument('--datapath', default='../../ppi', help="The data path.") parser.add_argument("--task_type", default="full", help="The node classification task type (full and semi). Only valid for cora, citeseer and pubmed dataset.") parser.add_argument("--no_tensorboard", default=False, help="Disable writing logs to tensorboard") parser.add_argument('--lradjust', action='store_true', default=False, help='Enable leraning rate adjust.(ReduceLROnPlateau or Linear Reduce)') OUTPUT_PATH = r'F:\\testGCNfig1\\' test_flag = True args = parser.parse_args() args.cuda = torch.cuda.is_available() # random seed setting np.random.seed(args.seed) torch.manual_seed(args.seed) if args.cuda: torch.cuda.manual_seed(args.seed) def get_lr(optimizer): for param_group in optimizer.param_groups: return param_group['lr'] def sparse_mx_to_torch_sparse_tensor(sparse_mx): sparse_mx = sparse_mx.tocoo().astype(np.float32) indices = torch.from_numpy( np.vstack((sparse_mx.row, sparse_mx.col)).astype(np.int64)) values = torch.from_numpy(sparse_mx.data) shape = torch.Size(sparse_mx.shape) return torch.sparse.FloatTensor(indices, values, shape) def preprocess_graph(adj): adj = sp.coo_matrix(adj) adj_ = adj + sp.eye(adj.shape[0]) # 邻接矩阵加入自身信息，adj = adj + I rowsum = np.array(adj_.sum(1)) degree_mat_inv_sqrt = sp.diags(np.power(rowsum, -0.5).flatten()) # 节点的度矩阵 adj_normalized = adj_.dot(degree_mat_inv_sqrt).transpose().dot(degree_mat_inv_sqrt).tocoo() # 正则化，D^{-0.5}(adj+I)D^{-0.5} return adj_normalized def train(model, data_loader): model.train() total_loss = total_examples = 0 for data in data_loader: adj = sp.csr_matrix( (np.ones(data.num_edges), (data.edge_index[0, :], data.edge_index[1, :])), shape=[data.num_nodes, data.num_nodes]) adj_nrom = preprocess_graph(adj) adj_nrom = sparse_mx_to_torch_sparse_tensor(adj_nrom) adj_nrom = adj_nrom.cuda() data_x = data.x.cuda() data_y = data.y.cuda() optimizer.zero_grad() loss = criterion(model(data_x, adj_nrom), data_y) loss.backward() optimizer.step() total_loss += loss.item() * data.num_nodes total_examples += data.num_nodes return total_loss / total_examples, get_lr(optimizer) def test(model, data_loader): model.eval() ys, preds = [], [] for data in data_loader: ys.append(data.y) adj = sp.csr_matrix( (np.ones(data.num_edges), (data.edge_index[0, :], data.edge_index[1, :])), shape=[data.num_nodes, data.num_nodes]) adj_nrom = preprocess_graph(adj) adj_nrom = sparse_mx_to_torch_sparse_tensor(adj_nrom) adj_nrom = adj_nrom.cuda() data_x = data.x.cuda() out = model(data_x, adj_nrom) preds.append((out > 0).float().cpu()) y, pred = torch.cat(ys, dim=0).numpy(), torch.cat(preds, dim=0).numpy() f1_score_out = f1_score(y, pred, average='micro') if pred.sum() > 0 else 0 return f1_score_out path = args.datapath pre_transform = T.Compose([T.GCNNorm(), T.ToSparseTensor()]) train_dataset = PPI(path, split='train', pre_transform=pre_transform) val_dataset = PPI(path, split='val', pre_transform=pre_transform) test_dataset = PPI(path, split='test', pre_transform=pre_transform) train_loader = DataLoader(train_dataset, batch_size=2, shuffle=True) val_loader = DataLoader(val_dataset, batch_size=2, shuffle=False) test_loader = DataLoader(test_dataset, batch_size=2, shuffle=False) if not test_flag: test_cases = [ # base {'num_layers':2, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, # test layer 1 {'num_layers':3, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':4, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':8, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':16, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':32, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, # test self loops 6 {'num_layers':2, 'add_self_loops':True, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':4, 'add_self_loops':True, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':8, 'add_self_loops':True, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':16, 'add_self_loops':True, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':32, 'add_self_loops':True, 'add_bn':False, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, # test batch norm 11 {'num_layers':2, 'add_self_loops':False, 'add_bn':True, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':4, 'add_self_loops':False, 'add_bn':True, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':8, 'add_self_loops':False, 'add_bn':True, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':16, 'add_self_loops':False, 'add_bn':True, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':32, 'add_self_loops':False, 'add_bn':True, 'use_pairnorm':'None', 'drop_edge':1.0, 'activation':'linear'}, # test use_pairnorm 16 {'num_layers':2, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'PN', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':4, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'PN', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':8, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'PN', 'drop_edge':1.0, 'activation':'linear'}, {'num_layers':16, 'add_self_loops':False, 'add_bn':False, 'use_pairnorm':'PN', 'drop_edge':