Paper Info
Title
Energy-Efficient Adaptive Machine Learning on IoT End-Nodes With Class-Dependent Confidence
Abstract
Energy-efficient machine learning models that can run directly on edge devices are of great interest in IoT applications, as they can reduce network pressure and response latency, and improve privacy. An effective way to obtain energy-efficiency with small accuracy drops is to sequentially execute a set of increasingly complex models, early-stopping the procedure for “easy” inputs that can be confidently classified by the smallest models. As a stopping criterion, current methods employ a single threshold on the output probabilities produced by each model. In this work, we show that such a criterion is sub-optimal for datasets that include classes of different complexity, and we demonstrate a more general approach based on per-classes thresholds. With experiments on a low-power end-node, we show that our method can significantly reduce the energy consumption compared to the single-threshold approach.
Year
DOI
Venue
2020
10.1109/ICECS49266.2020.9294863
2020 27th IEEE International Conference on Electronics, Circuits and Systems (ICECS)
Keywords
DocType
ISSN
IoT end-nodes,class-dependent confidence,energy-efficient machine learning,edge devices,network pressure,energy-efficiency,stopping criterion,per-classes thresholds,low-power end-node,energy consumption,single-threshold,energy-efficient adaptive machine learning,response latency,privacy,output probabilities
Conference
2020 27th IEEE International Conference on Electronics, Circuits and Systems (ICECS), 2020, pp. 1-4
ISBN
Citations 
PageRank 
978-1-7281-6045-0
0
0.34
References 
Authors
5
6
Name
Order
Citations
PageRank
Francesco Daghero141.86
Alessio Burrello266.01
daniele jahier pagliari32113.19
Luca Benini4131161188.49
Enrico Macii52405349.96
Massimo Poncino646057.48