ATM-Net: Adaptive Termination and Multi-Precision Neural Networks for Energy-Harvested Edge Intelligence

TL;DR

ATM-Net introduces an adaptive neural network architecture for energy-harvested IoT devices, dynamically adjusting precision and depth to optimize energy use while maintaining high accuracy.

Contribution

It presents a novel architecture combining adaptive termination and multi-precision computing, tailored for energy-harvested edge devices, with an energy-aware scheduler for optimal performance.

Findings

Achieves up to 96.93% accuracy on benchmark datasets.

Reduces power consumption by up to 87.5% with Q4 quantization.

Significantly improves power-delay product for tested models.

Abstract

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.