:param edge_dim: edge feature matrix number of dimension

限制节点度的最大值

Query-Key product(normalization)

实验初始化

Floyd-Warshall算法查询最短路径(BFS遍历图)

绘制参数表格

测试模型

:param max_in_degree: max in degree of nodes :param max_out_degree: max out degree of nodes :param node_dim: hidden dimension

:param max_path_distance: max pairwise distance between nodes

:param dim_in: node feature matrix input number of dimension :param dim_q: query node feature matrix input number dimension :

:param num_heads: number of attention heads :param dim_in: node feature matrix input number of dimension :param dim_q: query

:param node_dim: node feature matrix input number of dimension :param edge_dim: edge feature matrix input number of dimension

:param num_layers: number of Graphormer layers :param input_node_dim: input dimension of node features :param node_dim: hidde

:param x: node feature matrix :param edge_index: edge_index of graph (adjacency list) :return: torch.Tensor, node embeddings

:param x: node feature matrix :param paths: pairwise node paths :return: torch.Tensor, spatial Encoding matrix

:param x: node feature matrix :param edge_attr: edge feature matrix :param edge_paths: pairwise node paths in edge indexes

:param query: node feature matrix :param key: node feature matrix :param value: node feature matrix :param edge_attr:

:param x: node feature matrix :param edge_attr: edge feature matrix :param b: spatial Encoding matrix :param edge_pat

h′(l) = MHA(LN(h(l−1))) + h(l−1) h(l) = FFN(LN(h′(l))) + h′(l) :param x: node feature matrix :param edge_attr: edge

:param data: input graph of batch of graphs :return: torch.Tensor, output node embeddings