UW Information Technology

September 26, 2019

Machine learning helps UW meet “always-on” wireless connectivity

By Ignacio Lobos

When a biology lecturer noticed Poll Everywhere, a classroom response app, was failing to accept some of his students’ answers, he knew he had a serious problem.

To find out what was happening, he sought help from UW-IT and Academic Technologies. They leveraged machine learning, analytics and data-driven insights to pinpoint an issue with the wireless connectivity and fix the problem.

For David Morton, director of UW-IT’s Network & Telecommunications Design, this particular glitch represented something larger: “Our students have much higher expectations of technology: it just needs to work all the time.”

After all, their grades can depend on “always-on network connectivity,” he explained.

However, it is not just students who need secure and dependable wireless networks. Faculty and staff are increasingly relying on complex applications and smart devices.

“Maintaining reliable communications is critical to everything we’re doing,” Morton said. “So, we’re leveraging machine learning to improve our systems, and in turn improving the classroom experience for students and faculty.”

Artificial intelligence keeps Wi-Fi humming along

Morton’s team uses Aruba NetInsight, a cloud-based system that employs artificial intelligence, to help track the health of the UW wireless network. The system analyzes the entire network, identifies performance problems in real time, and offers recommendations on how to fix them. As it tracks performance at the UW — and at 11 other major universities that also use the application — it learns as it amasses useful data that helps all institutions with critical decisions, such as where to expand Wi-Fi.

The glitch in the biology lecturer’s classroom was indeed complex — when a wireless connection went down, it automatically switched some students to another connection, leaving their wireless devices in limbo as the switch took place, and their answers unrecorded.

“It would have taken us countless hours of engineering sleuthing to track the problem and create a solution to prevent it from happening again,” Morton said. “But with machine learning, we zeroed in on the issues much faster.”

Learn more: Did you know eduroam is the UW’s preferred Wi-Fi connection service for all students, faculty and staff? You can also use eduroam in 101 countries around the world. Learn how to configure your devices at IT Connect.