动态调整精度与深度，让边缘设备在低功耗下高效运行神经网络。
ATM-Net: Adaptive Termination and Multi-Precision Neural Networks for Energy-Harvested Edge Intelligence
- 根据能量情况自适应切换32/8/4位精度和网络层数。
- 在4位量化下功耗降低87.5%，准确率最高达96.93%。
- 适合太阳能等能量受限的物联网设备部署。
ATM-Net是一种专为能量采集型物联网设备设计的新型神经网络架构，结合自适应终止机制与多精度计算。它通过早期退出点动态调整计算精度（32/8/4位）和网络深度，以响应能量供应变化。一个能量感知的任务调度器优化了能效与准确率的权衡。在CIFAR-10、PlantVillage和TissueMNIST数据集上的实验表明，ATM-Net在使用Q4量化时，相比32位运算将功耗降低87.5%，准确率最高达到96.93%。DenseNet-121的功耗延迟积从13.6J降至0.141J，ResNet-18从10.3J降至0.106J，验证了其在能量采集系统中的适用性。
原文摘要 · Abstract (English)
ATM-Net is a novel neural network architecture tailored for energy-harvested IoT devices, integrating adaptive termination points with multi-precision computing. It dynamically adjusts computational precision (32/8/4-bit) and network depth based on energy availability via early exit points. An energy-aware task scheduler optimizes the energy-accuracy trade-off. Experiments on CIFAR-10, PlantVillage, and TissueMNIST show ATM-Net achieves up to 96.93% accuracy while reducing power consumption by 87.5% with Q4 quantization compared to 32-bit operations. The power-delay product improves from 13.6J to 0.141J for DenseNet-121 and from 10.3J to 0.106J for ResNet-18, demonstrating its suitability for energy-harvesting systems.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。