Searching Neural Architectures for Sensor Nodes on IoT Gateways
Garavagno, Ragusa, Frisoli et al.
This paper presents an automatic method for the design of Neural Networks (NNs) at the edge, enabling Machine Learning (ML) access even in privacy-sensitive Internet of Things (IoT) applications. The proposed method runs on IoT gateways and designs NNs for connected sensor nodes without sharing the collected data outside the local network, keeping the data in the site of collection. This approach has the potential to enable ML for Healthcare Internet of Things (HIoT) and Industrial Internet of Things (IIoT), designing hardware-friendly and custom NNs at the edge for personalized healthcare and advanced industrial services such as quality control, predictive maintenance, or fault diagnosis. By preventing data from being disclosed to cloud services, this method safeguards sensitive information, including industrial secrets and personal data. The outcomes of a thorough experimental session confirm that -- on the Visual Wake Words dataset -- the proposed approach can achieve state-of-the-art results by exploiting a search procedure that runs in less than 10 hours on the Raspberry Pi Zero 2.
academic
Searching Neural Architectures for Sensor Nodes on IoT Gateways
本文提出了一种在边缘设备上自动设计神经网络的方法,使机器学习能够应用于隐私敏感的物联网(IoT)应用中。该方法运行在IoT网关上,为连接的传感器节点设计神经网络,而无需将收集的数据分享到本地网络之外,数据始终保留在收集地点。这种方法有潜力为医疗物联网(HIoT)和工业物联网(IIoT)启用机器学习,在边缘设计硬件友好的定制神经网络,用于个性化医疗和先进工业服务。实验结果表明,在Visual Wake Words数据集上,该方法能够达到最先进的结果,搜索过程在Raspberry Pi Zero 2上运行时间少于10小时。
输入:ξ_MEM, ξ_RAM, ξ_Flash, ξ_MAC
输出:Ŝ_α
1. k ← 1, Ŝ_α ← ∅
2. repeat:
3. c ← 0
4. while A(k,c) is feasible:
5. Ŝ_α ← Ŝ_α ∪ (k,c)
6. c ← c + 1
7. k ← k + 1
8. until (k,0) is not feasible