def __init__(self, in_channel: int, hidden_size: int, nc: int,
max_nodes: int):
super().__init__()
self.gcn_1 = nn.DenseSAGEConv(in_channel, hidden_size)
self.gcn_2 = nn.DenseSAGEConv(hidden_size, hidden_size)
self.pooling = DiffPool(hidden_size, int(0.25 * max_nodes))
self.gcn_3 = nn.DenseSAGEConv(hidden_size, 2 * hidden_size)
self.gcn_4 = nn.DenseSAGEConv(2 * hidden_size, 2 * hidden_size)
self.classifier = torch.nn.Linear(2 * hidden_size, nc)
def __init__(self, in_channels, out_channels1, out_channels2,
out_channels3, out_channels4, out_channels5, out_channels6,
maxnodes):
super(GraphClass, self).__init__()
self.sage1 = tnn.DenseSAGEConv(in_channels,
out_channels1,
normalize=True)
self.sage2 = tnn.DenseSAGEConv(out_channels1,
out_channels2,
normalize=True)
self.poolit1 = nn.Linear(out_channels1, 50)
self.poolit2 = nn.Linear(out_channels3, 15)
self.sage3 = tnn.DenseSAGEConv(out_channels1,
out_channels3,
normalize=True)
#self.sage4=tnn.DenseSAGEConv(out_channels3,out_channels4,normalize=True)
self.sage5 = tnn.DenseSAGEConv(out_channels3,
out_channels5,
normalize=True)
self.tr1 = nn.Linear(out_channels5, out_channels6)
self.tr2 = nn.Linear(out_channels6, 1)
self.fin = nn.Linear(15, 1)
self.drop4 = torch.nn.Dropout(p=0.4)
self.drop3 = torch.nn.Dropout(p=0.3)
self.drop2 = torch.nn.Dropout(p=0.2)
## Batch Normalization
self.bano1 = nn.BatchNorm1d(num_features=maxnodes)
self.bano2 = nn.BatchNorm1d(num_features=maxnodes)
self.bano3 = nn.BatchNorm1d(num_features=50)
self.bano4 = nn.BatchNorm1d(num_features=50)
self.bano5 = nn.BatchNorm1d(num_features=15)
def build_conv_model(self, node_in_dim, node_out_dim):
return pyg_nn.DenseSAGEConv(node_in_dim,node_out_dim)
def __init__(self, in_channel, hidden_size):
super().__init__()
self.pooling = nn.DenseSAGEConv(in_channel, hidden_size)
def __init__(self,in_channels, out_channels1, out_channels2,out_channels3,out_channels4, out_channels5,out_channels6, dropout):
super(GraphAE, self).__init__()
#GAE.reset_parameters(self)
self.out_channels2=out_channels2
"""
Encoding
"""
### Encoding
"""
self.sage1=tnn.DenseGCNConv(in_channels,out_channels1)
self.sage2=tnn.DenseGCNConv(out_channels1,out_channels2)
self.sage3=tnn.DenseGCNConv(out_channels2,out_channels3)
self.sage4=tnn.DenseGCNConv(out_channels3,out_channels4)
self.sage5=tnn.DenseGCNConv(out_channels4,out_channels5)
"""
#DenseSAGEConv(in_channels, out_channels, normalize=False, bias=True)
self.sage1=tnn.DenseSAGEConv(in_channels,out_channels1,normalize=True)
self.sage2=tnn.DenseSAGEConv(out_channels1,out_channels2,normalize=True)
self.sage3=tnn.DenseSAGEConv(out_channels2,out_channels3,normalize=True)
#self.sage4=tnn.DenseSAGEConv(out_channels3,out_channels4,normalize=True)
##self.sage42=nn.Linear(out_channels3,out_channels4)
self.sage5=tnn.DenseSAGEConv(out_channels3,out_channels5,normalize=True)
##self.poolit1=tnn.DenseSAGEConv(out_channels2,400)
self.poolit1=nn.Linear(out_channels2,400)
self.poolit2=nn.Linear(out_channels3,16)
##self.poolit2=tnn.DenseSAGEConv(out_channels4,200)
#self.poolit3=tnn.DenseSAGEConv(out_channels5,10)
#self.tr1=nn.Linear(out_channels5,out_channels6)
self.tr2=nn.Linear(out_channels5,64)
self.tr2=nn.Linear(out_channels5,64)
self.rev2=nn.Linear(64,out_channels5)
"""
self.revsage1=tnn.DenseGCNConv(out_channels1,in_channels)
self.revsage2=tnn.DenseGCNConv(out_channels2,out_channels1)
self.revsage3=tnn.DenseGCNConv(out_channels3,out_channels2)
self.revsage4=tnn.DenseGCNConv(out_channels4,out_channels3)
self.revsage5=tnn.DenseGCNConv(out_channels5,out_channels4)
"""
self.revsage1=tnn.DenseSAGEConv(out_channels1,in_channels,normalize=False)
self.revsage2=tnn.DenseSAGEConv(out_channels2,out_channels1,normalize=False)
self.revsage3=tnn.DenseSAGEConv(out_channels3,out_channels2,normalize=False)
self.revsage5=tnn.DenseSAGEConv(out_channels5,out_channels3,normalize=False)
self.drop5=torch.nn.Dropout(p=0.5)
self.drop4=torch.nn.Dropout(p=0.4)
self.drop3=torch.nn.Dropout(p=0.3)
## Batch Normalization
self.bano1 = nn.BatchNorm1d(num_features=1000)
self.bano2 = nn.BatchNorm1d(num_features=1000)
self.bano3 = nn.BatchNorm1d(num_features=400)
self.bano4 = nn.BatchNorm1d(num_features=400)
self.bano5 = nn.BatchNorm1d(num_features=16)
self.bano6 = nn.BatchNorm1d(num_features=10)