MUMBAI, India, Sept. 14 -- Intellectual Property India has published a patent application (202641104731 A) filed by Vardhaman College Of Engineering on August 31, 2026, for A Genetic Algorithm Approach To Architecture Optimization In Self-Evolving Neural Networks.
Inventors include Ms. Farhana Begum; Mr. M. Yugandhar; Mr. Shobanbabu R. J.; Mr. Nirmal Keshari Swain; Mr. S. Satheesh Kumar; and Dr. Srinivasulu Gogula.
The application for the patent was published on September 11, 2026, under issue no. 37/2026.
Abstract: A Genetic Algorithm Approach to Architecture Optimization in Self-Evolving Neural Networks is the proposed invention. The proposed invention discloses a genetic algorithm-based framework for architecture optimisation in self-evolving neural networks using Non-dominated Sorting Genetic Algorithm II (NSGA-II). This framework can optimise multiple objectives at the same time, including prediction accuracy, computational complexity, inference time, memory consumption and model generalisation, and automatically search for the best neural network architectures. Candidate network architectures are first encoded as chromosomes of structural parameters such as hidden layers, neurones, activation functions, connectivity patterns and learning hyperparameters. An initial population is evaluated and NSGA-II performs non-dominated sorting, crowding-distance estimation, selection, crossover and mutation to evolve better architectures for subsequent generations, maintaining diversity of solutions. The refined architecture is continuously optimised through a self-evolving mechanism that adapts the network structure in response to newly available data and changing operational conditions without manual redesign. The proposed framework improves the convergence speed, scalability, robustness and resource efficiency with less computational overhead.
Disclaimer: Curated by HT Syndication.
