MUMBAI, India, Aug. 12 -- Intellectual Property India has published a patent application (202611075800 A) filed by Dr. Bajarang Prasad Mishra; Dr. P. Mohanraj; Dr. G. Chandrasekar; Avinash Kumar; Dr. Siji Jose Pulluparambil; Dr. Rohan Marotrao Ingle; Mrs. A. Praveena; Mr. A. Prasath; Dr. Sangita Deshmukh; Dr. Jai Ruby; Dr. S. Muthuselvan; and Dr. Sumit Kushwaha on June 18, 2026, for Self-Optimizing Learning System Using Continuous Model Adaptation Techniques.
Inventors include Dr. Bajarang Prasad Mishra; Dr. P. Mohanraj; Dr. G. Chandrasekar; Avinash Kumar; Dr. Siji Jose Pulluparambil; Dr. Rohan Marotrao Ingle; Mrs. A. Praveena; Mr. A. Prasath; Dr. Sangita Deshmukh; Dr. Jai Ruby; Dr. S. Muthuselvan; and Dr. Sumit Kushwaha.
The application for the patent was published on August 07, 2026, under issue no. 32/2026.
Abstract: The present invention discloses a self-optimizing learning system using continuous model adaptation techniques for improving the performance, accuracy, and reliability of machine learning models in dynamic data environments. Conventional machine learning systems are generally trained using fixed historical datasets and deployed without continuous adaptation. As data patterns change over time, such static models may suffer from reduced prediction accuracy, increased error rate, and poor decision reliability. The proposed invention overcomes these limitations by providing an integrated adaptive learning framework capable of monitoring incoming data, detecting data drift and concept drift, retraining models automatically, optimizing learning parameters, and validating updated model versions before deployment. The system comprises a data acquisition module, pre-processing module, deployed learning model, performance monitoring module, drift detection module, adaptive retraining module, self-optimization module, feedback learning module, model validation module, and version control module. Incoming data is continuously processed and evaluated to identify changes in feature distribution, prediction behavior, and model performance. When performance degradation or drift is detected, the system automatically selects recent data, combines it with relevant historical data, and retrains the model. The self-optimization module adjusts hyperparameters, learning rate, feature selection, and decision thresholds to improve overall performance. The invention further provides feedback-based improvement using prediction errors, user responses, expert validation, and real-world outcomes. Updated models are deployed only after successful validation against the existing model, while previous stable versions are retained for rollback. The system is suitable for cloud, edge, hybrid, and distributed environments, and may be applied in healthcare, finance, cybersecurity, industrial automation, smart education, recommendation systems, and Internet of Things applications.
Disclaimer: Curated by HT Syndication.
