Davoodi and Hu land NSF grant to bring deep neural networks to the internet of things

Azadeh Davoodi and Yu Hen Hu, both professors of electrical and computer engineering, have received a $499,867 grant from the National Science Foundation to investigate ways to synthesize complex deep neural networks (DNN) as a collection of smaller DNNs on distributed internet of things (IoT) devices.

Deep neural networks are a type of complex machine learning algorithm that are growing more sophisticated all the time. The algorithms are used or will be used to power smart cities, autonomous cars and advanced healthcare, among other applications. Typically, these networks gather data via sensors which send the information to the cloud for processing. However, DNNs cannot operate when the cloud is unavailable, and cloud-based networks may not be customizable to individual needs or environments.

This grant will aid Davoodi, who serves as principal investigator, and Hu, co-principal investigator, in exploring techniques to deploy deep neural networks on simple internet of things devices. These devices can then work in parallel as a network to perform the same inference tasks as DNNs based in the cloud. The hope is that these new networks of devices will enable faster development of smart IoT devices, paving the way to the next generation of smart services.