MUMBAI, India, May 1 -- Intellectual Property India has published a patent application (202641051637 A) filed by Vellore Institute Of Technology, Vellore, Tamil Nadu, on April 23, for 'a system and method for modular additive neural architecture enabling selective machine unlearning.'
Inventor(s) include Ramalingam Murugan; and Rajeswari C.
The application for the patent was published on May 1, under issue no. 18/2026.
According to the abstract released by the Intellectual Property India: "A system (100) and method for modular additive neural architecture enabling selective machine unlearning is disclosed. The system (100) comprises a data preprocessing component configured to transform input data, a partitioning component configured to divide the data into independent subsets, and a plurality of neural computational blocks configured to process respective subsets using independent parameter sets. Each neural computational block generates an output, and an aggregation component combines the outputs using an additive operation followed by an activation function. The system (100) further includes a selective unlearning component configured to identify a target data subset and perform retraining of a corresponding neural computational block while restricting parameter updates in other blocks. The architecture enables structural decomposition, independent parameterization, additive aggregation, and localized retraining within the system (100)."
Disclaimer: Curated by HT Syndication.
