← Search

Weiyan Wang

2 accepted papers

2024

Prometheus: Out-of-distribution Fluid Dynamics Modeling with Disentangled Graph ODE

ICML 2024poster

Fluid dynamics modeling has received extensive attention in the machine learning community. Although numerous graph neural network (GNN) approaches have been proposed for this problem, the problem of out-of-distribution (OOD) generalization remains underexplored. In this work, we propose a new large…

Cited by 10SourcePDFScholar
2024

Surge Phenomenon in Optimal Learning Rate and Batch Size Scaling

NeurIPS 2024poster

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 optimizers typically update model parameters using the sign of gradients, resulting in more stable convergence curves. The l…

Cited by 6SourcePDFScholar