PlatformX: An End-to-End Transferable Platform for Energy-Efficient Neural Architecture Search
Xiaolong Tu, Dawei Chen, Kyungtae Han, Onur Altintas, Haoxin Wang
TL;DR
PlatformX tackles the practical barriers of energy-aware HW-NAS for edge devices by unifying an energy-conscious search space, transferable kernel-level energy prediction, Pareto-based multi-objective optimization, and automated high-resolution on-device profiling. It expands traditional NAS spaces to reveal energy-efficient architectures, initializes predictors from synthetic kernel data, and adapts them across devices with minimal calibration. A gradient-guided Pareto search efficiently navigates trade-offs between energy and accuracy, validated by real-device measurements that continuously refine both the model pool and the predictors. The approach delivers substantial reductions in search time while preserving hardware-faithful energy estimates, enabling deployable, energy-efficient architectures that outperform MobileNet-V2 on edge hardware.
Abstract
Hardware-Aware Neural Architecture Search (HW-NAS) has emerged as a powerful tool for designing efficient deep neural networks (DNNs) tailored to edge devices. However, existing methods remain largely impractical for real-world deployment due to their high time cost, extensive manual profiling, and poor scalability across diverse hardware platforms with complex, device-specific energy behavior. In this paper, we present PlatformX, a fully automated and transferable HW-NAS framework designed to overcome these limitations. PlatformX integrates four key components: (i) an energy-driven search space that expands conventional NAS design by incorporating energy-critical configurations, enabling exploration of high-efficiency architectures; (ii) a transferable kernel-level energy predictor across devices and incrementally refined with minimal on-device samples; (iii) a Pareto-based multi-objective search algorithm that balances energy and accuracy to identify optimal trade-offs; and (iv) a high-resolution runtime energy profiling system that automates on-device power measurement using external monitors without human intervention. We evaluate PlatformX across multiple mobile platforms, showing that it significantly reduces search overhead while preserving accuracy and energy fidelity. It identifies models with up to 0.94 accuracy or as little as 0.16 mJ per inference, both outperforming MobileNet-V2 in accuracy and efficiency. Code and tutorials are available at github.com/amai-gsu/PlatformX.