On the Power of Gradual Network Alignment Using Dual-Perception Similarities

Network alignment (NA) is the task of finding the correspondence of nodes between two networks based on the network structure and node attributes. Our study is motivated by the fact that, since most of existing NA methods have attempted to discover all node pairs
at once
, they do not harness information enriched through
interim
discovery of node correspondences to more accurately find the next correspondences during the node matching. To tackle this challenge, we propose
${\sf Grad\text{-}Align}$
, a new NA method that
gradually
discovers node pairs by making full use of node pairs exhibiting strong consistency, which are easy to be discovered in the early stage of gradual matching. Specifically,
${\sf Grad\text{-}Align}$
first generates node embeddings of the two networks based on graph neural networks along with our
layer-wise reconstruction loss
, a loss built upon capturing the first-order and higher-order neighborhood structures. Then, nodes are gradually aligned by computing
dual-perception similarity
measures including the
multi-layer embedding similarity
as well as the
Tversky similarity
, an asymmetric set similarity using the Tversky index applicable to networks with different scales. Additionally, we incorporate an edge augmentation module into
${\sf Grad\text{-}Align}$
to reinforce the structural consistency. Through comprehensive experiments using real-world and synthetic datasets, we empirically demonstrate that
${\sf Grad\text{-}Align}$
consistently outperforms state-of-the-art NA methods.