基于加权元学习的节点分类算法

Node Classification Algorithm Based on Weighted Meta-Learning

受注意力机制和直推式学习方法的启发,提出一种基于加权元学习的节点分类算法.首先利用欧氏距离计算元学习子任务间数据分布的差异;然后利用子图的邻接矩阵计算捕获子任务间数据点的结构差异;最后将二者转化为权重对元训练阶段更新元学习器过程进行加权,构建优化的元学习模型,解决了经典元学习算法在元训练阶段所有元训练子任务的损失是等权重更新元学习器参数的问题.该算法在数据集Citeseer和Cora上的实验结果优于其他经典算法,证明了该算法在少样本节点分类任务上的有效性.

Inspired by attention mechanism and transductive learning method,we proposed a node classification algorithm based on weighted meta-learning.Firstly,Euclidean distance was used to calculate the difference of data distribution between subtasks in meta-learning.Secondly,adjacency matrices of subgraph was used to calculate and capture structural difference of data points between subtasks.Finally,the captured information above between subtasks were converted into weights to weight the process of updating the meta-learner in the meta-training procedure,and an optimized meta-learning model was constructed to solve the problem that the loss of all meta-training subtasks in meta-training procedure of classical meta-learning algorithms was equal-weight to update the parameters of meta-learners.The experimental results of this algorithm on Citeseer and Cora datasets are superior to other classical algorithms,which demonstrates the effectiveness of the algorithm on few-shot node classification task.