AI Thrust Seminar| Graph neural network enhanced air quality forecasting

While the rapid progress of urbanization has brought us great convenience in living, working, transportation, etc., it inevitably causes serious air pollution concerns world-widely. As reported by the World Health Organization (WHO), air pollution has become the world's largest environmental health risk and created a substantial burden on the economy in many developing countries. In fact, the potential losses caused by air pollution could be reduced or even prevented through effective pollution emission controls and early interventions, if we can accurately predict the air quality index (AQI) in each region of a city ahead of time. In this talk, I will share our recent work on fine-grained air quality forecasting by leveraging the recent advances in graph neural network, including (1) self-supervised hierarchical graph neural network for semi-supervised city-wide air quality forecasting, and (2) joint air quality and weather forecasting based on multi-adversarial spatiotemporal networks.