Summer Centered: Linyi Chen ’18 Trains A.I. to Improve Kidney Exchange Networks

In the complex landscape of organ donation, Linyi Chen ’18 is aiding an effort to use artificial intelligence (A.I.) as a means to efficiently match recipients with the donors they need. At the University of Maryland (UMD), Chen is joining the National Science Foundation’s (NSF) Research Experiences for Undergraduates program in the school’s computer science department to find kidney donor-recipient matches using neural networks, a type of A.I. She is supported by the Marian E. Koshland Integrated Natural Science Center (KINSC).

Commonly, someone who needs a kidney will enter the kidney exchange market with someone willing to donate, but is incompatible for a direct donation (such as an A-blood-type brother willing to donate to his B-blood-type sister who needs a kidney). In the exchange network, this incompatible pair can find another incompatible pair (such as a B-blood-type donor and an A-blood-type patient) to match with and create a compatible cycle of donation. Or, a chain of pairs can be identified so one pair’s donor can donate to another pair’s recipient, and that pair’s donor can donate to another recipient, and so on.

Chen’s research aims to use computational algorithms that “learn” from the exchange network’s data and identify the largest cycles or chains of donation that include the greatest number of pairs. In a large chain or cycle, more pairs can find someone to donate to them, and someone else to donate to.

“In the real world, the kidney market is changing everyday, and it is not always better to [match] existing cycles or chains immediately,” said Chen, a computer science and mathematics double major. “If we wait, maybe the cycle or [chain] could be larger if more compatible patient-donor pairs join the market later.”

In her 10 weeks of research, Chen is applying the kidney exchange network’s data to a neural network, a computational function that can study data sets in order to better predict and identify patterns or perform other functions.

“We are training our algorithms on real world data to ‘teach’ them how to think,” said Chen. “It is always exciting for us to train the neural net and see that after training it performs what we expect it to be.”

Since the research position is organized through the NSF, Chen, an international student from Shenzhen, China, wasn’t eligible for funding. Thanks to the KINSC, Chen is able to work within a discipline that inspires her.

After learning algorithmic and programming languages in her Haverford classes, Chen’s research gives her to opportunity to apply her knowledge to a real-life problem. Her experience working with AI this summer has opened the door to a possible thesis topic and even a career path that she’s passionate about.

“With AI, we can more easily to generalize useful information from a large set of data and to solve more real life problems,” she said. “This research experience helps me to find my interest in this field and I will apply to grad schools with this area of study.”

“Summer Centered” is a series exploring our students’ Center-funded summer work.