activity
20242026
collaborators
Showing cs.LGShow all

5 papers · 1 filter

cs.LG2025

Spatiotemporal Forecasting as Planning: A Model-Based Reinforcement Learning Approach with Generative World Models

Hao Wu, Yuan Gao, Xingjian Shi +9

To address the dual challenges of inherent stochasticity and non-differentiable metrics in physical spatiotemporal forecasting, we propose Spatiotemporal Forecasting as Planning (S…

cs.LG2025

BeamVQ: Beam Search with Vector Quantization to Mitigate Data Scarcity in Physical Spatiotemporal Forecasting

Weiyan Wang, Xingjian Shi, Ruiqi Shu +10

In practice, physical spatiotemporal forecasting can suffer from data scarcity, because collecting large-scale data is non-trivial, especially for extreme events. Hence, we propose…

cs.LG2024

Surge Phenomenon in Optimal Learning Rate and Batch Size Scaling

Shuaipeng Li, Penghao Zhao, Hailin Zhang +10

In current deep learning tasks, Adam style optimizers such as Adam, Adagrad, RMSProp, Adafactor, and Lion have been widely used as alternatives to SGD style optimizers. These optim…

cs.LG2024

BeamVQ: Aligning Space-Time Forecasting Model via Self-training on Physics-aware Metrics

Hao Wu, Xingjian Shi, Ziyue Huang +6

Data-driven deep learning has emerged as the new paradigm to model complex physical space-time systems. These data-driven methods learn patterns by optimizing statistical metrics a…

cs.LG2024

Spatio-Temporal Fluid Dynamics Modeling via Physical-Awareness and Parameter Diffusion Guidance

Hao Wu, Fan Xu, Yifan Duan +6

This paper proposes a two-stage framework named ST-PAD for spatio-temporal fluid dynamics modeling in the field of earth sciences, aiming to achieve high-precision simulation and p…