Machine Learning / Deep Learning / Optimization / Quantum Computing / Simulation / Data Assimilation
Technologies that create near-reality models by combining large-scale simulations with observational data
We research foundational technologies that fuse AI techniques—including machine learning and deep learning—with mathematical optimization and simulation to solve complex real-world problems. Our work spans large language model (LLM) applications, improving prediction model accuracy through data assimilation (AI prediction and data assimilation), AI surrogate models that rapidly approximate simulations, and mathematical approaches to combinatorial optimization.
We also advance diverse learning paradigms: knowledge distillation for transferring knowledge from teacher to lightweight models, robust learning methods for class-imbalanced data, and reinforcement learning for acquiring strategies through trial and error. These techniques serve as foundations directly linked to solving real-world challenges our lab addresses, including traffic optimization, weather prediction, disaster forecasting, and smart agriculture.
Keywords
