MUMBAI, India, July 7 -- Intellectual Property India has published a patent application (202641079115 A) filed by Amrita Vishwa Vidyapeetham on June 26, 2026, for A System And A Method For Training A Global Predictive Model In A Federated Learning Environment.
Inventors include Narasimhan, Harini; and Dora, Prajna.
The application for the patent was published on July 03, 2026, under issue no. 27/2026.
Abstract: The present invention relates to a system and method for training a global predictive model in a federated learning environment using a plurality of client devices. The method comprises selecting a subset of client devices, identifying source and target client devices, generating synthetic data to augment local datasets, receiving local model updates, and aggregating the updates to train the global model. The invention addresses challenges arising from non-identically and independently distributed (non-IID) data, including heterogeneity, imbalance, and skew, through adaptive client selection, synthetic data augmentation, and fairness-aware optimization, thereby improving model convergence, robustness, and predictive performance. Figure 1 is the representative figure.
Disclaimer: Curated by HT Syndication.
