Research Interests
My research connects generative modeling, continuous-time dynamics, and adaptive neural architectures, with applications in computer vision and graph-based spatiotemporal forecasting.
Generative AI
I study how generative models can use structural and conditional information to guide the generation process. My goal is to make generation more controllable and computationally efficient while preserving output quality and consistency.
Continuous-Time Modeling
I study how to learn the evolution of structured systems from sparse, irregular, and noisy observations. Using neural differential equations, I aim to capture temporal dynamics, interactions among entities, and uncertainty within continuous-time models.
Adaptive Neural Architectures
I study how neural networks can adapt their parameters and computational depth to individual inputs. My goal is to make computation more flexible, balancing predictive performance, parameter storage, and inference cost.
