About this deal
BatchNorm2d module with lazy initialization of the num_features argument of the BatchNorm2d that is inferred from the input. ConvTranspose2d module with lazy initialization of the in_channels argument of the ConvTranspose2d that is inferred from the input. The differentiable group normalization layer from the "Towards Deeper Graph Neural Networks with Differentiable Group Normalization" paper, which normalizes node features group-wise via a learnable soft cluster assignment. Applies the Softplus function Softplus ( x ) = 1 β ∗ log ( 1 + exp ( β ∗ x ) ) \text{Softplus}(x) = \frac{1}{\beta} * \log(1 + \exp(\beta * x)) Softplus ( x ) = β 1 ∗ lo g ( 1 + exp ( β ∗ x )) element-wise.
Conv1d module with lazy initialization of the in_channels argument of the Conv1d that is inferred from the input. The Adversarially Regularized Graph Auto-Encoder model from the "Adversarially Regularized Graph Autoencoder for Graph Embedding" paper. Applies Batch Normalization over a 5D input (a mini-batch of 3D inputs with additional channel dimension) as described in the paper Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift . Furthermore, an interesting discussion concerns the trade-off between representational power (usually gained through learnable functions implemented as neural networks) and the formal property of permutation invariance ( Buterez et al.Performs aggregations with one or more aggregators and combines aggregated results, as described in the "Principal Neighbourhood Aggregation for Graph Nets" and "Adaptive Filters and Aggregator Fusion for Efficient Graph Convolutions" papers. The Graph Auto-Encoder model from the "Variational Graph Auto-Encoders" paper based on user-defined encoder and decoder models. The Principal Neighbourhood Aggregation graph convolution operator from the "Principal Neighbourhood Aggregation for Graph Nets" paper. The Temporal Graph Network (TGN) memory model from the "Temporal Graph Networks for Deep Learning on Dynamic Graphs" paper. To facilitate further experimentation and unify the concepts of aggregation within GNNs across both MessagePassing and global readouts, we have made the concept of Aggregation a first-class principle in PyG.
The Weisfeiler Lehman (WL) operator from the "A Reduction of a Graph to a Canonical Form and an Algebra Arising During this Reduction" paper. During training, randomly zeroes some of the elements of the input tensor with probability p using samples from a Bernoulli distribution. Applies a multi-layer Elman RNN with tanh \tanh tanh or ReLU \text{ReLU} ReLU non-linearity to an input sequence.Aggregation functions play an important role in the message passing framework and the readout functions of Graph Neural Networks. The spline-based convolutional operator from the "SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernels" paper. InstanceNorm1d module with lazy initialization of the num_features argument of the InstanceNorm1d that is inferred from the input. InstanceNorm3d module with lazy initialization of the num_features argument of the InstanceNorm3d that is inferred from the input. Importantly, MultiAggregation provides various options to combine the outputs of its underlying aggegations ( e.