©作者 | Dream
单位 | 浙江大学
研究方向 | 图表示学习
论文链接:
Please kindly note that GSL, although conceptually related, is fundamentally distinct to related problems such as graph generation that target at generating novel, diversified graphs [Du et al., 2021], or graph learning that aims to recover the Laplacian matrix of a homophilous graph corresponding to given node features.
Direct approaches:将邻接矩阵看作是自由学习矩阵,通过 GNN 的参数来优化它们。