Spatiotemporal Learning for Urban Systems
sparse sensing, imputation, forecasting, and implicit neural representations for city-scale traffic data.
This thread focuses on learning from sparse, noisy, and incomplete urban observations. I have worked on low-rank tensor models, graph neural networks, Transformers, MLP-Mixers, and implicit neural representations for traffic imputation, estimation, and forecasting.
Representative work: ImputeFormer, STINR, C-LoRA, and Tensor4Kriging.