MUMBAI, India, Jan. 2 -- Intellectual Property India has published a patent application (202541122971 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on Dec. 5, 2025, for 'graph neural network system for categorical data imputation.'
Inventor(s) include Dr. Vijayakumar K; and Mr. S S Madavan Annamalai.
The application for the patent was published on Jan. 2, under issue no. 01/2026.
According to the abstract released by the Intellectual Property India: "The present disclosure provides a computer-implemented method (100) for imputing missing categorical data. The method (100) loads (102) a categorical dataset containing data records with categorical attributes, encodes (104) the categorical attributes into numerical representations using label encoding, constructs (108) a graph representation where data records are connected based on similarity relationships, trains (112) a Graph Attention Network model on the constructed graph to learn relationships between data records, and predicts (114) missing categorical values using the trained model by leveraging feature information and graph structure. Graph construction involves calculating Hamming distances between records, identifying k-nearest neighbors, and creating bidirectional edges between neighboring records. The Graph Attention Network incorporates multi-head attention mechanisms to dynamically weight neighboring node influence during imputation and uses cross-entropy loss for optimization."
Disclaimer: Curated by HT Syndication.
