Federated Prompt Learning for Weather Foundation Models on Devices
Shengchao Chen, Guodong Long, Tao Shen, Jing Jiang, Chengqi Zhang
TL;DR
This work tackles on-device weather forecasting under Federated Learning by addressing data heterogeneity across devices and data homogeneity within devices, while reducing communication load. It introduces FedPoD, a framework built on Adaptive Prompt Tuning and Dynamic Graph Modeling to personalize forecasts via lightweight prompts and server-side graph-based aggregation that prioritizes collaboration among geographically or distributionally similar clients. The approach yields state-of-the-art performance on real-world on-device weather datasets, with substantial parameter efficiency and privacy-preserving capabilities, as demonstrated through extensive ablations and analyses. The results suggest strong potential for scalable, privacy-conscious, on-device weather prediction, with future work aimed at long-horizon forecasting and broader spatiotemporal tasks.
Abstract
On-device intelligence for weather forecasting uses local deep learning models to analyze weather patterns without centralized cloud computing, holds significance for supporting human activates. Federated Learning is a promising solution for such forecasting by enabling collaborative model training without sharing raw data. However, it faces three main challenges that hinder its reliability: (1) data heterogeneity among devices due to geographic differences; (2) data homogeneity within individual devices and (3) communication overload from sending large model parameters for collaboration. To address these challenges, this paper propose Federated Prompt Learning for Weather Foundation Models on Devices (FedPoD), which enables devices to obtain highly customized models while maintaining communication efficiency. Concretely, our Adaptive Prompt Tuning leverages lightweight prompts guide frozen foundation model to generate more precise predictions, also conducts prompt-based multi-level communication to encourage multi-source knowledge fusion and regulate optimization. Additionally, Dynamic Graph Modeling constructs graphs from prompts, prioritizing collaborative training among devices with similar data distributions to against heterogeneity. Extensive experiments demonstrates FedPoD leads the performance among state-of-the-art baselines across various setting in real-world on-device weather forecasting datasets.