Will Deep Learning Accelerate the Spread of IoT?


With over 55 million people infected with COVID-19 worldwide, hospitals are under enormous strain, and remote care is needed more than ever. To meet the growing demand, the European Commission has created an (EUR) $8 million project called “IntellIoT”. It is one of several projects that will use Internet of Things (IoT) technology to boost the efficiency of hospitals and other organizations.

The IntellIoT project is a consortium of thirteen organizations that includes Philips, Siemens, the University of St. Gallen, EURECOM, Aalborg University, Sphynx Analytics, and University of Oulu.

The thirteen partners will test a number of potential IoT-based solutions for health monitoring and interventions, as well as large-scale medical data analysis. The goal will be to save hospitals both time and money while reducing the amount of in-person patient interactions that are more likely to endanger staff. 

For example, the University General Hospital of Heraklion in Greece will partner with Philips, a health technology company, to develop AI algorithms. They will create algorithms for use in diagnostic healthcare devices and sensors intended to speed diagnostics and enhance accuracy. Likewise, they will test new IoT-equipped technologies to serve as go-betweens for patients and medical staff during remote patient care. 

“In the beginning of the pandemic we all faced a lack of information and data, be it patients, doctors, and political decision makers,” Prof. Fragkiskos Parthenakis, a doctor at the hospital, told Forbes. “Intelligent IoT solutions that provide humanised, trusted and secure data will help facilitate the use of distributed AI for decision making and better service in healthcare in the future.”

MIT Researchers Develop “Deep Learning” System For IoT

Major advances are expected to bring deep learning to IoT machines that learn from mimicking the human brain. For example, machine learning algorithms provide Google users with better search results. They can also curate targeted content for social media users. 

Currently, IoT devices run on small computer chips called “microcontrollers.” Because microcontrollers have minimal processing abilities, data needs to be sent to the cloud for deep learning analysis. This makes IoT devices vulnerable to security breaches. 

To solve this problem, MIT researchers developed a system of tightly-knit neural networks for IoT devices. Dubbed “MCUNet,” the system can run deep learning algorithms locally, which means IoT devices don’t need to send data to the cloud.

MCUNet consists of two components: a “TinyEngine” and a “TinyNAS.” The TinyEngine serves as an inference engine that dictates resource management. The TinyNAS, a neural architecture search method, determines a neural network structure for the TinyEngine to run on. The TinyNAS is customizable, meaning it can create compact neural networks with the best potential outcome for a microcontroller. The TinyEngine then produces the necessary code for TinyNAS’ customized neural network. This “tiny deep learning” system eliminates unnecessary code in order to make IoT devices more nimble and efficient. 

As IoT devices continue to proliferate, the improved efficiency and security benefits of MCUNets are likely to accelerate their spread.

Guide to the Internet of Things

In the IEEE Guide to the Internet of Things, an eight-course program, participants learn about IOT-applications, challenges, and future opportunities. This program is designed for professionals working in engineering, IT, computer science, and related fields across all industries.

Connect with an IEEE Content Specialist to receive a custom quote for your organization today.

Interested in the course for yourself? Visit the IEEE Learning Network!


Chandler, Simon. (5 November 2020). How The Internet Of Things Can Help Hospitals Cope With Coronavirus. Forbes.

Ackerman, Daniel. (13 November 2020). System brings deep learning to “internet of things” devices. MIT News Office.

, , , ,

No comments yet.

Leave a Reply