Gated Graph Sequence Neural Networks

| arxiv | ICLR 2016 |

where \(σ(i(h_v^{(T)}, x_v))\) acts as a soft attention mechanism that decides which nodes are relevant to the current graph-level task. \(i\) and \(j\) are neural networks that take the concatenation of \(h_v^{(T)}\) and \(x_v\) as input and outputs real-valued vectors. The \(tanh\) functions can also be replaced with the identity.