MUMBAI, India, Jan. 9 -- Intellectual Property India has published a patent application (202541134067 A) filed by Bvrit Hyderabad College Of Engineering For Women; Manjula Pilaka; and Dr Aneesh Sreevallabh Chivukula, Hyderabad, Telangana, on Dec. 31, 2025, for 'kolmogorov-arnold network-based residual learning system for forecasting government bond yields.'
Inventor(s) include Bvrit College Engineering For Women; Manjula Pilaka; Dr Aneesh Sreevallabh Chivukula; Dr K Vsunitha; and Dr Aruna Rao S L.
The application for the patent was published on Jan. 9, under issue no. 02/2026.
According to the abstract released by the Intellectual Property India: "A Kolmogorov-Arnold Network (KAN)-based residual learning system for forecasting Government bond yields is disclosed. Historical yields and derived features are arranged into sliding windows and first processed by a base model, such as an LSTM or LSTM ensemble, to generate primary forecasts in the original yield space. The same windows are flattened and used to train a compact KAN to model forecast residuals, which are added to the base predictions at inference to obtain final yields. The modular design allows the KAN block to attach to different base models, providing a parameter efficient way to capture complex non-linear residual structure and support more accurate and interpretable bond-yield forecasts. At inference time, the system outputs final forecasts as the sum of the base model prediction and the KAN residual correction. The architecture provides a compact, modular, and accurate residual-correction mechanism that improves bond-yield forecasts compared to base models alone."
Disclaimer: Curated by HT Syndication.
