AutoFed: Personalized Federated Traffic Prediction via Adaptive Prompt

2026-04-21 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Robotics & Autonomous Systems · Depth: Expert, extended

Summary

AutoFed is a novel Personalized Federated Learning (PFL) framework designed for traffic prediction, addressing privacy concerns and the non-IID (non-independent and identically distributed) data problem among clients. Unlike traditional PFL methods that require extensive manual hyper-parameter tuning for graph feature engineering and network architecture, AutoFed eliminates this need. It features a Personalized Predictor (PP) that adapts to local data and a Federated Representor (FR) that distills local data into a compact, globally shared prompt matrix using a client-aligned adapter. This prompt then guides the personalized predictor, enabling cross-client knowledge sharing while maintaining local data specificity. Experiments on real-world Travel Demand Prediction (TDP) and Traffic Flow Prediction (TFP) datasets demonstrate AutoFed's superior performance across diverse scenarios, achieving state-of-the-art results in TDP and comparable performance in TFP with reduced communication costs and faster convergence.

Key takeaway

For AI Scientists and Machine Learning Engineers developing federated learning solutions for traffic prediction, AutoFed offers a significant advantage by eliminating manual hyper-parameter tuning. Your teams can achieve superior performance and faster convergence without the need for prior knowledge of client or dataset configurations, which often impedes practical deployment. Consider integrating AutoFed's prompt-learning approach to streamline development and enhance model robustness in privacy-sensitive, non-IID traffic data environments.

Key insights

AutoFed offers manual-free personalized federated traffic prediction by using a global prompt to guide local models.

Principles

• Adaptive graph structures remove manual design needs.
• Auto-encoders can autonomously learn stable patterns.
• Shared linear layers with distinct BN layers align features.

Method

AutoFed uses an AE-denoiser for robust feature extraction, a graph time series encoder for compression, and a client-aligned adapter with FedBN to generate a global prompt matrix for a personalized predictor's decoder.

In practice

• Employ adaptive graph models like AGCRN for varied traffic networks.
• Use AE-based denoisers to extract stable patterns from noisy time series.
• Implement FedBN in FL for non-IID data alignment.

Topics

• AutoFed
• Personalized Federated Learning
• Traffic Prediction
• Prompt Learning
• Adaptive Graph Convolutional Recurrent Network

Code references

• RS2002/AutoFed
• lichuan210/FedTPS

Best for: AI Scientist, Machine Learning Engineer, Research Scientist

Related on AIssential

• Federated Nested Learning: Collaborative Training of Self-Referential Memories for Test-Time Adaptation — FedNL enables federated models to learn adaptive memory update rules, not just static weights, for robust test-time adaptation.
• Improving Crash Frequency Prediction from Simulated Traffic Conflicts Using Machine Learning Based Microsimulation — ML-based microsimulation improves traffic conflict realism for crash frequency prediction.
• pFedUL: Layer-Aware Federated Unlearning for Personalized Federated Learning — Layer-aware federated unlearning balances data removal completeness with personalization preservation in pFL.
• Auto-FP: An Experimental Study of Automated Feature Preprocessing for Tabular Data — Automating feature preprocessing for tabular data can be framed as HPO or NAS, with evolution-based algorithms performing best.
• C2FL: Clustered Continual Federated Learning under Spatial and Temporal Drift — C2FL enables robust federated learning in mobile, privacy-sensitive systems by combining spatial clustering with temporal drift adaptation.
• Black Forest Labs' new Self-Flow technique makes training multimodal AI models 2.8x more efficient — Black Forest Labs has introduced Self-Flow, a novel self-supervised flow matching framework that significantly enhances the efficiency and capability of training multimodal AI models by eliminating re…

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.