This invention presents a novel framework for predictive modeling of semantic drift using Koopman operators, transforming non-linear changes in word meaning into a linear system within a lifted space. By leveraging polynomial observables and Dynamic Mode Decomposition (DMD), the invention provides accurate long-term...

We propose SpectralONN, a novel spectral-domain neural network architecture that operates directly on Fourier-transformed signal representations using learnable complex-valued operators. Unlike traditional spatial neural networks, SpectralONN encodes both magnitude and phase dynamics within a structured, mathematica...

This invention introduces C3QG, a novel transformer-based system for context-controlled, explainable, and efficient question generation. Leveraging advanced evaluation metrics (BERTScore, METEOR), SHAP-based interpretability, and GPU-accelerated fine-tuning of FLAN-T5 transformers, the invention achieves significant...