nn

Accessor([nodes, activation, index])	Accessor is a neural network module that provides access to specific tensor data with optional activation.
Concatenator(activation[, nodes])	This class is designed to concatenate multiple tensors along the last dimension and apply an activation function to the concatenated result.
Dirichlet(activation, dirichlet_name[, nodes])	Dirichlet is a neural network module that overwrites values with that of dirichlet field.
Contraction(activation[, nodes])	Contraction is a neural network module that performs a contraction operation on the input tensor.
DeepSets(lambda_config, gamma_config, ...[, ...])	DeepSets is a neural network module that performs permutation invariant / equivariant operation on the input tensor.
EnEquivariantMLP(nodes[, activations, ...])	EnEquivariantMLP is a neural network module that performs an E(n)-equivariant operation on the input tensor.
EnEquivariantTCN(nodes, kernel_sizes, ...[, ...])	EnEquivariantTCN is a neural network module that performs an E(n)-equivariant operation on the time series input tensors.
Einsum(activation, equation, **kwards)	Einsum is a neural network module that performs a einsum operation on the input tensors.
GCN(nodes, support_name[, activations, ...])	GCN is a neural network module that performs a graph convolution operation on the input tensor.
Identity([nodes])	Identity is a neural network module that returns the input tensor.
IsoGCN(nodes, isoam_names, propagations, ...)	IsoGCN is a neural network module that performs an spatial pifferential operation on the input tensor.
MLP(nodes[, activations, dropouts, bias])	Multi Layer Perceptron
PInvMLP(reference_name, **kwards)	Pseudo inverse of the reference MLP layer.
Proportional(nodes)	Proportional, i.e., strictly linear, layer
Rearrange(pattern, axes_lengths, ...)	Rearrange is a neural network module that performs a rearrange operation on the input tensor.
Reducer(activation, operator[, nodes])	Reducer is a neural network module that performs a reduction operation on the input tensor.
Share(reference_name, **kwards)	Share module have same operations as the reference module.
SimilarityEquivariantMLP(nodes, scale_names)	Similarity-equivariant Multi Layer Perceptron as in https://proceedings.mlr.press/v235/horie24a.html.
SPMM(factor, support_name[, transpose])	Sparse Matrix Multiplication Network
TCN(nodes, kernel_sizes, dilations, activations)	Temporal Convolutional Networks
TimeSeriesToFeatures([nodes, activation])	TimeSeriesToFeatures is a neural network module that converts a time series tensor into a non time series tensor by summing the time series tensor along the time dimension.