MUMBAI, India, July 30 -- Intellectual Property India has published a patent application (202641087009 A) filed by Srm University- Ap on July 16, 2026, for System And Method For Graph Neural Network Based Link Prediction Using Node Influence And Structural Similarity.
Inventors include Baswani, Madhusudhana Rao; Yarra, Khyathisree; Dr. Prasanthi Boyapati; Dr. Tangirala Jaya Lakshmi; and Dr. Y V Nandini.
The application for the patent was published on July 24, 2026, under issue no. 30/2026.
Abstract: A system (100) for graph neural network based link prediction comprises processors (102) and a memory (104) storing instructions that cause the system (100) to receive graph data representing a network comprising nodes and edges, compute node centrality values for each node based on network topology, and compute structural similarity scores for candidate node pairs based on neighborhood information. The system (100) generates structural prior scores by combining aggregated centrality values derived from node centrality values with structural similarity scores using a fusion parameter controlling relative contributions. The system (100) generates node embeddings using a graph neural network encoder performing message passing operations across the network, and estimates link probabilities for candidate node pairs using a structure-aware decoder based on the node embeddings and structural prior scores.
Disclaimer: Curated by HT Syndication.
