"""
cgc_classification model is named after the structure of the graph neural network.
The graph neural network is structured with a convolutional , graph attention,
and another convolutional layer. The cgc_classificatin model was the model
tested int the publication Introducing CARATE: Finally speaking chemistry.
:author: Julian M. Kleber
"""
import torch
import torch.nn.functional as F
from torch.nn import Linear
from torch_geometric.nn import global_add_pool, GraphConv, GATConv
import numpy as np
import sklearn.metrics as metrics
import logging
from carate.models.base_model import Model
logger = logging.getLogger(__name__)
logging.basicConfig(
filename="carate.log",
encoding="utf-8",
level=logging.DEBUG,
format="%(asctime)s %(message)s",
)
[docs]
class Net(Model):
"""
The Net is the core algorithm and needs a constructor and a
forward pass. The train, test and evaluation methods are implemented
in the evaluation module with the Evaluation class.
"""
def __init__(
self,
dim: int,
num_features: int,
num_classes: int,
num_heads: int = 16,
dropout_gat: float = 0.6,
dropout_forward=0.5,
*args,
**kwargs,
) -> None:
super(Net, self).__init__(
dim=dim, num_classes=num_classes, num_features=num_features
)
self.num_heads = num_heads
self.dropout_gat = dropout_gat
self.dropout_forward = dropout_forward
self.gat1 = GraphConv(self.dim * 16, self.dim)
self.conv1 = GraphConv(self.num_features, self.dim)
self.conv2 = GATConv(self.dim, self.dim, dropout=0.6, heads=self.num_heads)
self.fc1 = Linear(self.dim, self.dim)
self.fc2 = Linear(self.dim, self.num_classes)
[docs]
def forward(self, x, edge_index, batch, edge_weight=None):
x = F.relu(self.conv1(x, edge_index, edge_weight))
x = F.dropout(x, p=0.5, training=self.training)
x = F.relu(self.conv3(x, edge_index, edge_weight))
x = F.relu(self.conv5(x, edge_index, edge_weight))
x = global_add_pool(x, batch)
x = F.relu(self.fc1(x))
x = F.dropout(x, p=0.5, training=self.training)
x = self.fc2(x)
return x
def __str__(self):
return "gcc_classification"