Carbon-Aware Continuous Learning for Sustainable Real-Time Machine Learning Analytics
Abstract: Real-time machine learning (ML) analytics models deployed on edge servers often experience degraded inference accuracy due to data drift. Continuous learning mitigates this by periodically retraining models using newly collected data. However, retraining incurs significant computational over-head, increasing energy consumption and carbon footprint several-fold. Existing approaches for reducing carbon foot-print predominantly focus on general workload scheduling strategies, such as shifting jobs to periods or regions with lower carbon intensity. However, they neglect continuous-learning-specific parameters like retraining cadence, data subsampling rates, and teacher model capacity. Consequently, these approaches miss opportunities to further reduce the carbon footprint and enhance inference accuracy in continuous learning systems under dynamic data drift. To address this gap, it is essential to jointly optimize these continuous-learning-specific parameters while simultaneously determining optimal retraining timings based on real-time fluctuations in carbon intensity and accuracy degradation. In this pa-per, we propose a novel Carbon-footprint-aware Continuous Learning (CCL) scheme that minimizes carbon emissions during model retraining without sacrificing inference accuracy. Distinct from prior workload scheduling approaches, CCL adaptively adjusts retraining cadence, data subsampling rate, and model capacity based on predictive models that forecast data drift severity and carbon intensity dynamics. Our adaptive real-time optimization approach consistently achieves a near-optimal balance between accuracy and carbon footprint under dynamic conditions. Experimental results demonstrate that CCL reduces carbon footprint by up to 54.1% compared to state-of-the-art carbon-agnostic methods, with negligible accuracy degradation.
