MUMBAI, India, Aug. 24 -- Intellectual Property India has published a patent application (202641097156 A) filed by Ms. G. Venkata Lakshmi; and Dr. G. Devadasu on August 12, 2026, for System And Method For Drift-Aware Continual Learning Using Dual-Memory Data Retention And Causality-Preserving Model Update Constraints.
Inventors include Dr. R Venkateswara Reddy; Mr. N. Eleswara Rao; Ms. U. Hema Latha; Ms. T. Lavanya; Mr. G. Uday Kishore; and Mr. P. Sai Kumar.
The application for the patent was published on August 21, 2026, under issue no. 34/2026.
Abstract: A system (100) and method for drift-aware continual learning using dual-memory data retention and causality-preserving model update constraints are disclosed. The system (100) includes a memory (102), a processor (104), and a communication module (106) configured to receive streaming input data and historical training information associated with machine learning environments. A drift detection module (108) identifies distributional and behavioral variations in incoming data. A dual- memory management module (110) maintains short-term and long-term memory regions for adaptive knowledge retention. A causal structure extraction module (114) generates causal relationship representations associated with model features and predictions. A causality-preserving constraint generation module (116) determines update constraints for restricting modification of causally significant model parameters during incremental learning. A continual learning optimization module (118) updates an operational machine learning model using replay information while reducing catastrophic forgetting and preserving causal consistency under evolving data conditions. Fig. 1
Disclaimer: Curated by HT Syndication.
