MUMBAI, India, Aug. 17 -- Intellectual Property India has published a patent application (202641095074 A) filed by Cmr Institute Of Technology; Cmr Technical Campus; and Cmr College Of Engineering & Technology on August 06, 2026, for A System And Method For Cross-Domain Transfer Learning Using Meta-Representation Alignment Networks.
Inventors include Dr Mummala Muni Babu; Sanuvala Ganga; Ramesh Azmeera; J Rekha; Mr. D. Parvateeswara Rao; and Ms. V. Manga.
The application for the patent was published on August 14, 2026, under issue no. 33/2026.
Abstract: A system and method for cross-domain transfer learning using meta-representation alignment networks are disclosed. The system receives source domain data and target domain data having heterogeneous feature distributions and encodes the data into latent feature representations using a feature encoding module. A meta- representation learning module generates domain-invariant meta-representations that capture transferable knowledge across domains. A meta-representation alignment network aligns the meta-representations in a shared representation space by minimizing a domain discrepancy measure. A predictive model trained on the source domain is adapted to the target domain based on the aligned meta-representations. The system outputs predictions for the target domain with improved generalization performance, particularly in low-resource or limited labeled data scenarios. The disclosed approach enables efficient knowledge transfer across domains and enhances robustness of machine learning models in diverse application environments.
Disclaimer: Curated by HT Syndication.
