Fig. 2
From: Predicting circRNA-drug sensitivity associations via graph attention auto-encoder
The process of using two-layer GATE to reconstruct the features of node 3. the neighbors of the node 3 are (1, 2, 3, 4, 5). we note that \(h^{(0)}_i=x_i, h_i=h^{(2)}_i={\hat{h}}^{(2)}_i\) , and \({\hat{x}}_i={\hat{h}}^{(0)}_i\), \(\forall i \in {1,2,\ldots ,N}\)