MUMBAI, India, July 24 -- Intellectual Property India has published a patent application (202641085269 A) filed by D Sandhya Rani; and Dr. A. Raji Reddy on July 12, 2026, for System And Method For Memory-Aware Federated Learning Using Causality- Driven Data Selection And Adaptive Model Synchronization.
Inventors include R. Ravi; A. Ganapathi; K. Madhu; G. Pavan; S. Raghavendra; and Shafana Bakshi.
The application for the patent was published on July 17, 2026, under issue no. 29/2026.
Abstract: A system (100) and method for memory-aware federated learning using causality-driven data selection and adaptive model synchronization are disclosed. The system (100) includes a memory (102), a processor (104), and a communication module (106) configured to support distributed federated learning across one or more resource-constrained client devices. A local data profiling module (108) analyzes locally available training data to identify redundancy characteristics, feature relevance, and memory utilization conditions. A causal relevance determination module (110) generates causal relevance scores for training samples using causal dependency analysis. A memory-aware prioritization module (112) selectively retains or suppresses data according to causal significance and device memory thresholds. A federated training control module (114) performs local model training using prioritized data subsets, while an adaptive synchronization module (116) dynamically modifies synchronization operations according to memory availability, causal contribution metrics, and convergence conditions. FIG. 1
Disclaimer: Curated by HT Syndication.
