Towards Stream Learning on Embedded Systems: Benchmarking the Memory Consumption of Stream Learning Methods
2026-08-31 • Machine Learning
Machine LearningArtificial IntelligencePerformance
AI summaryⓘ
The authors studied how different stream learning methods use memory and processing power over time, especially when working with limited resources like sensors. They tested seven classifiers on many data streams with strict memory limits and found that some models start too large or keep growing, causing them to fail on small devices. Compact methods work well on tiny budgets, but larger budgets let other methods perform better. They suggest future research should focus on controlling resource use alongside adapting to data changes.
stream learningconcept driftresource usageembedded systemsadaptive ensemblesincremental treesHoeffding Treesmodel sizememory budgetlatency
Authors
Sebastian Buschjäger, Nuwan Gunasekara, Heitor Murilo Gomes
Abstract
Stream learning is commonly evaluated through predictive performance and adaptation to concept drift. However, sustained operation of a stream learner also requires predictable and bounded resource usage even on long streams. This requirement becomes even more critical when learning moves from servers to near-sensor embedded systems where memory and processing are scarce resources. In state-of-the-art stream learning, however, we perceive a strong focus on concept drift adaptation, whereas resource usage is often an evaluation byproduct. To close this gap, we benchmark seven representative stream classifiers on 13 real and synthetic streams under model-size budgets from 128\,KiB to approximately 8\,MiB. Our benchmark comprises a total of 6,463 experiments. We measure failure-aware accuracy, peak model size, time to budget exhaustion, and prediction-plus-update latency. The results reveal two distinct resource failure modes. Adaptive ensembles can exceed small budgets almost immediately because of their initial footprint, even when their size remains stable thereafter. Incremental trees can fit initially but grow throughout a long stream, with HoeffdingTrees (HT) and Extremely Fast Decision Trees (EFDT) increasing by median factors of 7.37 and 5.87. Explicitly compact methods remain the only viable option under the smallest budgets, but are usually overtaken as larger budgets make adaptive ensembles competitive. Hence, many state-of-the-art methods are only partially applicable in embedded systems or for long-running systems. We therefore call on the stream-learning community to make bounded resource usage a first-class design objective alongside drift adaptation, and propose concrete steps toward this goal, including an API through which stream learners can explicitly expose and respect resource budgets.