Four CS Undergraduates Earn CRA Research Recognition

Mathematics and Computer Science senior Reed Oei was among five students nationwide selected for a 2021 Computing Research Association (CRA) Outstanding Undergraduate Researcher Runner-Up Award. Fellow CS students Rittika Adhikari, Nathan Ju, and Xiangchen Song received Honorable Mention status for the CRA Award, which recognizes undergraduate students in North American universities who show outstanding research potential in an area of computing research. 

Oei, who worked with CS Professors Darko Marinov’s and Tao Xie’s groups for two years, has made fundamental contributions to the detection and fixing of flaky tests—a type of software test that produces inconsistent results, making it hard for software developers to know if their program is functioning properly.

“Flaky tests have become a big concern for companies like Apple, Facebook, Google, and others because finding and fixing them is difficult and expensive,” said Oei, who co-developed the tools iDFlakies and iFixFlakies to find and fix flaky tests in Java code. “This work is important because it allows software developers to have confidence in their test suites, despite containing flaky tests.”

Oei’s flaky test work resulted in three published papers and appeared in the top ACM and IEEE conferences on software engineering and testing.

In a second area of research, Oei is working with researchers at Carnegie Mellon University on developing programming languages for writing smart contracts, which are programs that run on blockchains and are vital to managing cryptocurrencies or online auctions. Specifically, Oei improved the Obsidian smart contract programming language and designed a new language called Psamathe; he published the results of this work in four papers.

Oei is currently working with Illinois Mathematics Associate Professor Philipp Hieronymi on the creation of an automatic theorem prover called Pecan, which he’s using to prove many theorems about Sturmian words. Oei plans to attend graduate school in the fall and study programming languages and logic so he can pursue a career in academia.

“It’s nice to have the recognition because a lot of people apply for this award and they’re doing excellent research, too,” Oei said. “I’m grateful to all my mentors for having taught me how to do research successfully.”

Honorable Mention recipients

Working with Assistant Professor Sanmi Koyejo’s group, CS senior Rittika Adhikari is examining the lack of trustworthiness in modern machine learning (ML) models, which are used to facilitate driverless cars, stock market prediction trends, and medical diagnoses. Her work aims to build interpretability tools that accurately and effectively explain how these models make their decisions.

In a related research project, Adhikari is developing an ML model to assist doctors in diagnosing Covid-19 through chest X-rays. Her approach uses federated learning, a distributed ML technique used to train an algorithm across multiple decentralized devices with local data samples without sharing them.

CS senior Nathan Ju’s primary research efforts focus on quantum complexity theory. In the summer of 2020, he worked with researchers at the University of Waterloo investigating whether noisy quantum circuits solve problems that are hard for comparable classical computers. 

Ju discovered a clever proof that helped explain the separation between noisy low-depth quantum circuits and low-depth classical circuits. His work proved that noisy intermediate-scale quantum (NISQ) computers—devices with imperfect and few qubits—have an advantage over classical computers, which may ultimately help unlock the power of NISQ devices. He presented a talk based on this work at QIP 2021, the top conference in theoretical quantum information research.

Mathematics and Computer Science student Xiangchen Song (BS '20), who has been working in CS Professor Jiawei Han’s group, conducts research at the intersection of graph mining and text mining and their applications in bioinformatics.

In the area of heterogeneous graph mining, he has developed a random projection-based graph embedding learning method that analyzes the expert finding problem. In the area of text mining, he proposed the Triplet Matching Network for taxonomy completion task, which proposed a novel task and outperforms existing methods. In the area of text-rich graph mining, Song proposed a novel graph convolutional network (GCN) architecture that performs joint convolution on text feature and graph feature; this work also has applications on e-commerce searching problems.

Song has published  his work at several international conferences such as ISMB, AAAI, and WSDM.