Deep Generative Models for Spatiotemporal Trajectory

Description

The availability of realistic human movements, commonly referred to as trajectories, is of crucial importance in various application domains such as urban planning, transportation, and public health. For instance, in public health, understanding the spread of an epidemic relies on accurate knowledge of human movements. However, obtaining real-world trajectories is challenging due to concerns around privacy and commercial sensitivities, which limits their availability. Consequently, generating synthetic but realistic trajectories has emerged as a critical research area. This project aims to study a novel deep learning framework that can incorporate different reasonable inductive biases (e.g., distance , semantic meanings) to generate with black-box deep neural networks.


People


Haowen Lin

CS PhD Student, USC