Tag
The paper proposes PROLA, a principal-orthogonal low-rank adaptation method that allocates a fixed rank budget between pretrained weights' principal directions and their orthogonal complement, and introduces a Predictive Spatiotemporal Weather Downscaling task where PROLA outperforms existing low-rank adaptation baselines across diverse datasets.
This paper conducts a controlled study on advection-aware graph nowcasting for distributed solar ramp forecasting, finding that accurate cloud-motion features are as important as graph structure and introducing a self-supervised estimator that reduces forecast error.
This paper presents a latent dynamical model using a heart-rate-aware neural ODE and graph-based mesh autoencoder to model full-cycle ventricular motion from cine cardiac MRI. Applied to 72,386 UK Biobank participants, the model improves heart failure risk prediction over conventional cardiac markers.
Proposes a unified risk map modeling framework for autonomous driving that integrates traffic flow and collision risks in partially observable environments, using spatiotemporal modeling and diffusion-based scenario generation. Outperforms state-of-the-art occlusion-aware baselines on the Waymo Open Motion Dataset.