Source code for topomodelx.nn.simplicial.scone_layer

"""Simplicial Complex Net Layer."""
from typing import Literal

import torch

from topomodelx.base.aggregation import Aggregation


[docs] class SCoNeLayer(torch.nn.Module): """Implementation of the SCoNe layer proposed in [1]_. Parameters ---------- in_channels : int Input dimension of features on each edge. out_channels : int Output dimension of features on each edge. update_func : Literal['relu', 'sigmoid', 'tanh'] Update function to use when updating edge features. Notes ----- This is the architecture proposed for trajectory prediction on simplicial complexes. For the trajectory prediction architecture proposed in [1]_, these layers are stacked before applying the boundary map from 1-chains to 0-chains. Finally, one can apply the softmax operator on the neighbouring nodes of the last node in the given trajectory to predict the next node. When implemented like this, we get a map from (ordered) 1-chains (trajectories) to the neighbouring nodes of the last node in the 1-chain. References ---------- .. [1] Roddenberry, Mitchell, Glaze. Principled simplicial neural networks for trajectory prediction. ICML 2021. https://proceedings.mlr.press/v139/roddenberry21a.html .. [2] Papillon, Sanborn, Hajij, Miolane. Equations of topological neural networks (2023). https://github.com/awesome-tnns/awesome-tnns/ .. [3] Papillon, Sanborn, Hajij, Miolane. Architectures of topological deep learning: a survey on topological neural networks (2023). https://arxiv.org/abs/2304.10031. """ def __init__( self, in_channels: int, out_channels: int, update_func: Literal["relu", "sigmoid", "tanh"] = "tanh", ) -> None: super().__init__() self.in_channels = in_channels self.out_channels = out_channels self.weight_0 = torch.nn.parameter.Parameter( torch.Tensor(self.in_channels, self.out_channels) ) self.weight_1 = torch.nn.parameter.Parameter( torch.Tensor(self.in_channels, self.out_channels) ) self.weight_2 = torch.nn.parameter.Parameter( torch.Tensor(self.in_channels, self.out_channels) ) self.aggr_on_edges = Aggregation("sum", update_func)
[docs] def reset_parameters(self, gain: float = 1.0) -> None: """Reset learnable parameters.""" torch.nn.init.xavier_uniform_(self.weight_0, gain=gain) torch.nn.init.xavier_uniform_(self.weight_1, gain=gain) torch.nn.init.xavier_uniform_(self.weight_2, gain=gain)
[docs] def forward( self, x: torch.Tensor, incidence_1: torch.Tensor, incidence_2: torch.Tensor ) -> torch.Tensor: r"""Forward pass. The forward pass was initially proposed in [1]_. Its equations are given in [2]_ and graphically illustrated in [3]_. .. math:: \begin{align*} &🟥 \quad m^{(1 \rightarrow 0 \rightarrow 1)}_{y \rightarrow \{z\} \rightarrow x} = (L_{\downarrow,1})_{xy} \cdot h_y^{t,(1)} \cdot \Theta^{t,(1 \rightarrow 0 \rightarrow 1)}\\ &🟥 \quad m_{x \rightarrow x}^{(1 \rightarrow 1)} = h_x^{t,(1)} \cdot \Theta^{t,(1 \rightarrow 1)}\\ &🟥 \quad m_{y \rightarrow \{z\} \rightarrow x}^{(1 \rightarrow 2 \rightarrow 1)} = (L_{\uparrow,1})_{xy} \cdot h_y^{t,(1)} \cdot \Theta^{t,(1 \rightarrow 2 \rightarrow 1)}\\ &🟧 \quad m_{x}^{(1 \rightarrow 0 \rightarrow 1)} = \sum_{y \in \mathcal{L}_\downarrow(x)} m_{y \rightarrow \{z\} \rightarrow x}^{(1 \rightarrow 0 \rightarrow 1)}\\ &🟧 \quad m_{x}^{(1 \rightarrow 2 \rightarrow 1)} = \sum_{y \in \mathcal{L}_\uparrow(x)} m_{y \rightarrow \{z\} \rightarrow x}^{(1 \rightarrow 2 \rightarrow 1)}\\ &🟩 \quad m_x^{(1)} = m_{x}^{(1 \rightarrow 0 \rightarrow 1)} + m_{x \rightarrow x}^{(1 \rightarrow 1)} + m_{x}^{(1 \rightarrow 2 \rightarrow 1)}\\ &🟦 \quad h_x^{t,(1)} = \sigma(m_x^{(1)}) \end{align*} Parameters ---------- x: torch.Tensor, shape = (n_edges, in_channels) Input features on the edges of the simplicial complex. incidence_1 : torch.sparse, shape = (n_nodes, n_edges) Incidence matrix :math:`B_1` mapping edges to nodes. incidence_2 : torch.sparse, shape = (n_edges, n_triangles) Incidence matrix :math:`B_2` mapping triangles to edges. Returns ------- torch.Tensor, shape = (n_edges, out_channels) Output features on the edges of the simplicial complex. """ z1 = incidence_2 @ incidence_2.T @ x @ self.weight_2 z2 = x @ self.weight_1 z3 = incidence_1.T @ incidence_1 @ x @ self.weight_0 return self.aggr_on_edges([z1, z2, z3])