Reproduce “An Energy-Aware Approach to Design Self-Adaptive AI-based Applications on the Edge” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Alessandro Tundo, Marco Mobilio, Shashikant Ilager, Ivona Brandić, Ezio Bartocci, Leonardo Mariani
A methodology is introduced for designing self-adaptive AI applications on edge devices that dynamically balance performance objectives with energy consumption. An empirical evaluation on a pedestrian detection system shows the approach significantly improves energy efficiency by up to 81% and processing rates by up to 233% compared to static configurations, while achieving comparable or superior accuracy through intelligent mode switching.
Key insights
- Empirical measurement of objectives on target hardware is essential for accurate multi-objective optimization in edge AI applications, as simulations are often inaccurate.
- Meta-heuristic search can identify high-quality, optimal trade-off configurations with significantly reduced exploration time (e.g., 10x faster) compared to near-exhaustive search.
- Self-adaptive AI applications, capable of dynamically switching operational modes, achieve superior overall objective balancing (accuracy, efficiency, energy) compared to static configurations in dynamic environments.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.