9/27/2022 Aaron Seidlitz, Illinois CS
With guidance from their advisors – professors Deepak Vasisht and Gagandeep Singh – both Liu and Xu remain committed to finding better application of their work in building robust machine learning wireless systems.
Written by Aaron Seidlitz, Illinois CS
At the end of August, two Illinois Computer Science PhD students, Zikun Liu and Calvin Xu, found out they had earned inclusion as one of just 19 winners for the 2022 Qualcomm Innovation Fellowship (North America).
Their project, entitled “Provably Robust Machine Learning (ML) for Wireless Systems,” will be guided by their advisors – Illinois CS professors Deepak Vasisht and Gagandeep Singh.
“We know that the Qualcomm Innovation Fellowship is very prestigious, and it’s a great encouragement and honor,” said Liu, a second-year PhD student advised by Vasisht. “It provides us an opportunity to show that our research impacts trustworthy machine learning wireless systems for real-world applications. Winning this fellowship also increases the exposure for this project within the industry, as we gain precious opportunities to work with fantastic research scientists from Qualcomm.”
Xu echoed Liu, regarding the importance of this fellowship opening their work up to industry collaborators and research scientists.
He mentioned that the topic – provably robust machine learning – has existed for a number of years, but it hasn’t made much of an impact in applied domains.
“Our research group is constantly looking for new application areas for trustworthy ML. Qualcomm is an industry leader in wireless systems. This fellowship was the perfect place to propose our idea as they solicited proposals about securing ML in wireless systems,” said Xu, a second-year PhD student advised by Singh.
The two students have an interest in different but complimentary research areas.
Liu’s work on adversarial attacks through his wireless machine learning algorithm, FIRE, combines well will Xu’s work on robust machine learning over two years with the MIT Lincoln Lab. Additionally, this type of work fits perfectly under the tutelage of Singh and Vasisht, who have spent years studying these domains and delivering results.
Throughout the process of earning this fellowship, Liu said the most memorable part was a Zoom meeting during the finalists’ round with leaders from Qualcomm.
“Calvin and I worked together to answer questions from a group of experienced researchers with Qualcomm. That represented a great opportunity for us, and I’m thrilled and humbled to have won the fellowship following that process,” Liu said. “It also demonstrates how well prepared we were, as both of us received amazing feedback prior to the finalists’ stage from our advisors and groupmates – without whom we could never have achieved this.”
The recognition earned through this fellowship also speaks directly to Xu’s motivation for studying machine learning at Illinois CS.
“Growing up, I loved participating in math contests, which helped me build up problem solving skills that flowed naturally into CS. Machine learning is my favorite part of CS because I also get to leverage the math skills that I've honed over the years,” Xu said. “Specifically, robust and adversarial machine learning are interesting as trust and verifiability are crucial for the widespread adoption of machine learning.”