Node Importance Estimation with Multiview Contrastive Representation Learning

Table 1

Table of symbols.

Symbol	Definition
	The input feature vector matrix of nodes
	The scaled adjacency matrix of the graph
, ,	Trainable weight matrixes
	Activation function
	Feature of the node or edge
	Neighboring nodes of node
	Score function that measures the similarity between two embeddings
	Indicator function
	Cross-view negative samples
,	Embeddings of node generated by first view and the second view
	Attention weight of node to node
,	Predicted scores of node of node view and node-edge interaction view
	Set of nodes with known importance scores
	Output node embedding matrix of GCN
	with added self-loops
	Learnable weight vector
||	Concatenation operator
	Embedding of node i in the layer k of GAT
	Number of nodes in the graph
	Fully connected neural network
	Hyperparameter
	Intraview negative samples
	Predicate between nodes and
	Valid ground truth importance value of node
	Predicted score of node