layer_agnn_conv: AGNNConv

In rdinnager/rspektral: What the Package Does (One Line, Title Case)

Description

\loadmathjax

An Attention-based Graph Neural Network (AGNN) as presented by Thekumparampil et al. (2018).

Mode: single, disjoint.

This layer expects a sparse adjacency matrix.

This layer computes: \mjdeqn Z = PX where \mjdeqn P_ij = \frac \exp \left( \beta \cos \left( X_i, X_j \right) \right) \sum\limits_k \in \mathcalN(i) \cup { i } \exp \left( \beta \cos \left( X_i, X_k \right) \right) and \mjeqn\beta is a trainable parameter.

Input

• Node features of shape (N, F);

• Binary adjacency matrix of shape (N, N).

Output

• Node features with the same shape of the input.

Usage

layer_agnn_conv(object, trainable = TRUE, activation = NULL, ...)

Arguments

trainable	boolean, if True, then beta is a trainable parameter. Otherwise, beta is fixed to 1
activation	activation function to use

rdinnager/rspektral documentation built on June 12, 2021, 1:26 a.m.