KAN4Met
In recent years we have witnessed two ground-breaking contributions in mathematical aspects of Deep Learning and Deep Learning applied to weather forecasting:
- KAN: Kolmogorov-Arnold Networks. Original paper published by Liu et al. in the arXiv on April 30 2024, currently cited 2000+ times.
- GraphCast: A Graph Neural Network for Weather Forecasting. Original paper published by Lam et al. in Science on November 14 2023, currently cited 1600+ times.
This proposed project (KAN4Met) aims to explore the intersection between the two fields opened by these fundamental contributions. In addition, its goal is to produce a ready-to-use ML model for weather prediction on the Iberian peninsula, in collaboration with the Spanish Weather Agency (AEMET). The project also aims to apply enhanced weather prediction models to the problem of route optimization in maritime transport, continuing the research developed by the IP of this proposal in the past three years.