Sebastian Buschjäger

Ensemble methods for stream mining necessitate managing multiple models and updating them as data distributions evolve. Considering the calls for more sustainability, established methods are however not sufficiently considerate of ensemble members’ computational expenses and instead overly focus on predictive capabilities. To address these challenges and enable green online learning, we propose heterogeneous online ensembles (HEROS). For every training step, HEROS chooses a subset of models from a pool of models initialized with diverse hyperparameter choices under resource constraints to train. We introduce a Markov decision process to theoretically capture the trade-offs between predictive performance and sustainability constraints. Based on this framework, we present different policies for choosing which models to train on incoming data. Most notably, we propose the novel Zeta-policy, which focuses on training near-optimal models at reduced costs. Using a stochastic model, we theoretically prove that our Zeta-policy achieves near optimal performance while using fewer resources compared to the best performing policy. In our experiments across 11 benchmark datasets, we find empiric evidence that our Zeta-policy is a strong contribution to the state-of-the-art, demonstrating highly accurate performance, in some cases even outperforming competitors, and simultaneously being much more resource-friendly.

We introduce AnytimeYOLO, a family of variants of the YOLO architecture that enables anytime object detection. Our AnytimeYOLO networks allow for interruptible inference, i.e., they provide a prediction at any point in time, a property desirable for safety-critical real-time applications. We present structured explorations to modify the YOLO architecture, enabling early termination to obtain intermediate results. We focus on providing fine-grained control through high granularity of available termination points. First, we formalize Anytime Models as a special class of prediction models that offer anytime predictions. Then, we discuss a novel transposed variant of the YOLO architecture, that changes the architecture to enable better early predictions and greater freedom for the order of processing stages. Finally, we propose two optimization algorithms that, given an anytime model, can be used to determine the optimal exit execution order and the optimal subset of early-exits to select for deployment in low-resource environments. We evaluate the anytime performance and trade-offs of design choices, proposing a new anytime quality metric for this purpose. In particular, we also discuss key challenges for anytime inference that currently make its deployment costly.

We introduce Splitting Stump Forests—small ensembles of weak learners extracted from a trained random forest. The high memory consumption of random forests renders them unfit for resource-constrained devices. We show empirically that we can significantly reduce the model size and inference time by selecting nodes that evenly split the arriving training data and applying a linear model on the resulting representation. Our extensive empirical evaluation indicates that Splitting Stump Forests outperform random forests and state-of-the-art compression methods on memory-limited embedded devices.

As machine learning models become increasingly integrated into various applications, the need for resource-aware deployment strategies becomes paramount. One promising approach for optimizing resource consumption is rejection ensembles. Rejection ensembles combine a small model deployed to an edge device with a large model deployed in the cloud with a rejector tasked to determine the most suitable model for a given input. Due to its novelty, existing research predominantly focuses on ad-hoc ensemble design, lacking a thorough understanding of rejector optimization and deployment strategies. This paper addresses this research gap by presenting a theoretical investigation into rejection ensembles and proposing a novel algorithm for training and deploying rejectors based on these novel insights. We give precise conditions of when a good rejector can improve the ensemble’s overall performance beyond the big model’s performance and when a bad rejector can make the ensemble worse than the small model. Second, we show that even the perfect rejector can overuse its budget for using the big model during deployment. Based on these insights, we propose to ignore any budget constraints during training but introduce additional safeguards during deployment. Experimental evaluation on 8 different datasets from various domains demonstrates the efficacy of our novel rejection ensembles outperforming existing approaches. Moreover, compared to standalone large model inference, we highlight the energy efficiency gains during deployment on a Nvidia Jetson AGX board.