Deep Learning for Urban Mobility Flows - New arXiv Papers

July 27, 2025

Our lab just published two new arXiv papers on deep learning for mobility flows in cities: 'UrbanPulse' uses graph and transformer models for highly fine-grained predictions across cities. 'FloGAN' uses conditional GANs to generate scenario-based mobility networks.

 

UrbanPulse: Prediction results of urban population flow at a single 15-minute interval for Los Angeles.

UrbanPulse: https://www.arxiv.org/abs/2507.17924
Authors: Hongrong Yang, Markus Schläpfer

FloGAN: https://arxiv.org/abs/2507.12053
Authors: Seanglidet Yean, Jiazu Zhou, Bu-Sung Lee, Markus Schläpfer

 

 

Columbia Affiliations
Department of Civil Engineering and Engineering Mechanics
School of Engineering and Applied Science