GNN汇集邻域信息代码

def _region_aggregate(self, feats, edge_dict):
    """
    feats: node features [n,feature_dim]
    edge_dict: adjacency list [[第一个node 邻接的node的编号],[],[]]  也就是node2node
    """
    N = feats.size()[0]
    pooled_feats = torch.stack([torch.mean(feats[edge_dict[i]], dim=0) for i in range(N)])

    return pooled_feats