MUMBAI, India, Nov. 2 -- Intellectual Property India has published a patent application (202611115788 A) filed by Dr. Promila Bahadur on September 28, 2026, for A Computer-Implemented System And Method For Imbalance-Aware Feature Processing Using Dual- Decoder Attention And Class-Aware Optimization.
Inventors include Dr. Promila Bahadur; Mr. Ishan Dwivedi; Yashasav Khandelwal; Mr. Pranjal Dwivedi; and Mr. Krishna Kumar Burman.
The application for the patent was published on October 30, 2026, under issue no. 44/2026.
Abstract: A computer-implemented artificial intelligence system (100) and method for processing class-imbalanced input data is disclosed. The system (100) comprises a data ingestion module (102), preprocessing module (104), encoder (106), global decoder (108), detail decoder (110), adaptive cross-attention fusion module (112), targeted class-aware sampling engine (114), hybrid weighted-loss optimization engine (116), and inference module (118). The encoder (106) generates hierarchical feature representations, which are separately processed to generate global and detail feature representations. The adaptive cross-attention fusion module (112) generates an attention-aligned fused feature representation. The targeted class-aware sampling engine (114) increases minority-class representation during training, while the hybrid weighted-loss optimization engine (116) combines multiple loss components to optimize the system (100) for generating minority-sensitive class predictions.
Disclaimer: Curated by HT Syndication.
