"""
GCN implementation of the DANN algorithm from `"Domain-Adversarial Training of Neural Networks"
<https://www.jmlr.org/papers/volume17/15-239/15-239.pdf>`_ paper
"""
from typing import Tuple
import torch
import torch.nn as nn
from torch import Tensor
from torch.autograd import Function
from GOOD import register
from GOOD.utils.config_reader import Union, CommonArgs, Munch
from .BaseGNN import GNNBasic
from .Classifiers import Classifier
from .GCNs import GCNFeatExtractor
[docs]@register.model_register
class DANN_GCN(GNNBasic):
r"""
The Graph Neural Network modified from the `"Domain-Adversarial Training of Neural Networks"
<https://www.jmlr.org/papers/volume17/15-239/15-239.pdf>`_ paper and `"Semi-supervised Classification with Graph Convolutional Networks"
<https://arxiv.org/abs/1609.02907>`_ paper.
Args:
config (Union[CommonArgs, Munch]): munchified dictionary of args (:obj:`config.model.dim_hidden`, :obj:`config.model.model_layer`, :obj:`config.dataset.dim_node`, :obj:`config.dataset.num_classes`, :obj:`config.dataset.num_envs`)
"""
def __init__(self, config: Union[CommonArgs, Munch]):
super().__init__(config)
self.feat_encoder = GCNFeatExtractor(config)
self.classifier = Classifier(config)
self.graph_repr = None
self.dc = nn.Linear(config.model.dim_hidden, config.dataset.num_envs)
self.config = config
[docs] def forward(self, *args, **kwargs) -> Tuple[torch.Tensor, torch.Tensor]:
r"""
The DANN-GCN model implementation.
Args:
*args (list): argument list for the use of arguments_read. Refer to :func:`arguments_read <GOOD.networks.models.BaseGNN.GNNBasic.arguments_read>`
**kwargs (dict): key word arguments for the use of arguments_read. Refer to :func:`arguments_read <GOOD.networks.models.BaseGNN.GNNBasic.arguments_read>`
Returns (Tensor):
[label predictions, domain predictions]
"""
out_readout = self.feat_encoder(*args, **kwargs)
dc_out = GradientReverseLayerF.apply(out_readout, self.config.train.alpha)
dc_out = self.dc(dc_out)
out = self.classifier(out_readout)
return out, dc_out
[docs]class GradientReverseLayerF(Function):
r"""
Gradient reverse layer for DANN algorithm.
"""
[docs] @staticmethod
def forward(ctx, x, alpha):
r"""
gradient forward propagation
Args:
ctx (object): object of the GradientReverseLayerF class
x (Tensor): feature representations
alpha (float): the GRL learning rate
Returns (Tensor):
feature representations
"""
ctx.alpha = alpha
return x.view_as(x) # * alpha
[docs] @staticmethod
def backward(ctx, grad_output):
r"""
gradient backpropagation step
Args:
ctx (object): object of the GradientReverseLayerF class
grad_output (Tensor): raw backpropagation gradient
Returns (Tensor):
backpropagation gradient
"""
output = grad_output.neg() * ctx.alpha
return output, None