A Two-Stage Data Selection Framework for Data-Efficient Model Training on Edge Devices

2025-05-22

AI Summary

Model training on edge devices is hindered by low throughput, limited storage, and heterogeneous data importance, resulting in poor data utilization. To address this, we propose Titan, a two-stage dynamic data selection framework. Titan introduces a novel synergistic mechanism combining coarse-grained pre-screening with fine-grained optimal selection; theoretically derives and implements a fine-grained importance-aware greedy selection strategy; and designs an online importance estimation module coupled with a lightweight modeling component, enabling zero-resource-conflict pipelining of data selection and model training via asynchronous execution. Evaluated on real-edge hardware, Titan reduces training time by 43%, improves final model accuracy by 6.2%, and incurs only negligible overhead in latency, memory footprint, and energy consumption.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search
• Humans and AI: Human-in-the-loop Machine Learning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

The demand for machine learning (ML) model training on edge devices is escalating due to data privacy and personalized service needs. However, we observe that current on-device model training is hampered by the under-utilization of on-device data, due to low training throughput, limited storage and diverse data importance. To improve data resource utilization, we propose a two-stage data selection framework {sf Titan} to select the most important data batch from streaming data for model training with guaranteed efficiency and effectiveness. Specifically, in the first stage, {sf Titan} filters out a candidate dataset with potentially high importance in a coarse-grained manner.In the second stage of fine-grained selection, we propose a theoretically optimal data selection strategy to identify the data batch with the highest model performance improvement to current training round. To further enhance time-and-resource efficiency, {sf Titan} leverages a pipeline to co-execute data selection and model training, and avoids resource conflicts by exploiting idle computing resources. We evaluate {sf Titan} on real-world edge devices and three representative edge computing tasks with diverse models and data modalities. Empirical results demonstrate that {sf Titan} achieves up to $43%$ reduction in training time and $6.2%$ increase in final accuracy with minor system overhead, such as data processing delay, memory footprint and energy consumption.

Problem

• Improves on-device ML training data utilization
• Selects high-importance data from streaming sources
• Reduces training time while boosting model accuracy

Innovation

• Two-stage data selection for edge training
• Pipeline co-execution of selection and training
• Optimal data batch identification strategy