MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202611060033 A) filed by Mohammed I. Habelalmateen; and Dr. Rohit Sharma on May 12, 2026, for System And Method For Adaptive Neural Network Optimization Using Dynamic Hyperparameter Tuning Mechanisms.
Inventors include Mohammed I. Habelalmateen; and Dr. Rohit Sharma.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: A system for adaptive neural network optimization using dynamic hyperparameter tuning mechanisms is disclosed. The system comprises a neural computation processor including tensor execution circuits configured for performing neural propagation operations associated with a neural network architecture. A telemetry acquisition unit continuously acquires runtime operational parameters comprising propagated gradient values, convergence trajectory measurements, processor utilization values, activation distribution measurements, memory occupancy values, communication latency measurements, and thermal accumulation measurements generated during neural execution operations. A convergence analysis processor generates multidimensional convergence representations corresponding to neural parameter evolution and identifies instability conditions including gradient explosion conditions, gradient vanishing conditions, activation saturation conditions, and oscillatory convergence conditions. A hyperparameter control processor dynamically modifies learning rate parameters, regularization coefficients, normalization scaling values, dropout allocation values, activation threshold values, and optimizer transition parameters during runtime neural training execution.
Disclaimer: Curated by HT Syndication.
