Continual Error Correction on Low-Resource Devices

2025-03-31

ACM SIGMM Conference on Multimedia Systems

AI Summary

To address the challenge of efficiently and continuously correcting AI model prediction errors on low-power devices, this paper proposes an edge-device collaborative few-shot continual error correction framework. Methodologically, it integrates server-side knowledge distillation with device-side dynamic prototype updating—eliminating the need for full-model retraining—and innovatively transfers foundation model knowledge to lightweight architectures, replacing parameter updates with prototype-based classification to drastically reduce on-device computational and memory overhead. The framework supports both image classification and object detection; on Food-101 and Flowers-102 under single-shot settings, it achieves over 50% error correction rate, less than 0.02% forgetting rate, and negligible inference overhead—validated via a real-world Android application. Its core contribution is the first demonstration of highly robust, low-forgetting online error correction under extreme resource constraints, simultaneously ensuring accuracy, efficiency, and practical deployability.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Large Vision Models
• Search and Optimization: Learning to Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The proliferation of AI models in everyday devices has highlighted a critical challenge: prediction errors that degrade user experience. While existing solutions focus on error detection, they rarely provide efficient correction mechanisms, especially for resource-constrained devices. We present a novel system enabling users to correct AI misclassifications through few-shot learning, requiring minimal computational resources and storage. Our approach combines server-side foundation model training with on-device prototype-based classification, enabling efficient error correction through prototype updates rather than model retraining. The system consists of two key components: (1) a server-side pipeline that leverages knowledge distillation to transfer robust feature representations from foundation models to device-compatible architectures, and (2) a device-side mechanism that enables ultra-efficient error correction through prototype adaptation. We demonstrate our system's effectiveness on both image classification and object detection tasks, achieving over 50% error correction in one-shot scenarios on Food-101 and Flowers-102 datasets while maintaining minimal forgetting (less than 0.02%) and negligible computational overhead. Our implementation, validated through an Android demonstration app, proves the system's practicality in real-world scenarios.

Problem

• Enables efficient AI error correction on low-resource devices
• Uses few-shot learning to correct misclassifications with minimal resources
• Combines server training and on-device prototypes for lightweight updates

Innovation

• Few-shot learning for AI error correction
• Server-side knowledge distillation to device models
• On-device prototype adaptation for efficient updates