Students' fresh perspectives lead ASU researcher to success

February 14, 2019

Huan Liu has built a renowned research career in the areas of social computing, data mining and artificial intelligence by letting his doctoral students lead the way.

“I think young people are more active, more creative,” said Liu, a professor of computer science and engineering in Arizona State University’s Ira A. Fulton Schools of Engineering. “But I can help them with my experience.” Professor Huan Liu works with his students in the Data Mining and Machine Learning Lab Arizona State University Professor Huan Liu's strategy of letting his doctoral students guide work in his research lab has helped him build a successful career in social computing, data mining and artificial intelligence. Pictured (from left): computer science graduate student Kai Shu, Huan Liu and computer science graduate student Deepak Mahudeswaran. Photo by Erika Gronek/ASU Download Full Image

As an AI researcher, Liu’s expertise focuses on discovering actionable patterns or insights from data, particularly social media data — a challenging type of information to work with.

His work attracts high-achieving students because, as users of social media, they know which problems they can solve, including tackling online issues such as fake news, cyberbullying, data privacy, thwarting malicious users and many more.

Together, Liu and his students are connecting the dots in social media data and making strides in improving our digital lives in a number of ways via AI.

From two to many

Liu takes an unconventional approach to working with doctoral students. When they join the School of Computing, Informatics, and Decision Systems Engineering, one of the six Fulton Schools, students choose from two general starting points for their doctoral research: feature selection for data mining or social computing.

From there, it’s up to each student to come up with a particular focus and make it their own. Whatever they end up with, Liu lets students take their projects with them into high-achieving careers in academia and industry while he starts all over again on a different focus with each new doctoral student.

Through this dynamic approach, Liu is building an influential and diverse body of research in these two areas.

Improving life online with feature selection and social computing

While the general public may view social media as a massive data source, Liu and his students use the concept of feature selection to make social media data even bigger and take advantage of unique characteristics of social media to make thin data thicker.

Feature selection sounds like narrowing the amount of data a researcher uses, but it is really about bringing out salient features that might be hidden in a data set. This means researchers have even more data to work with, and it’s data that is especially useful. It’s an effective tool in a data-mining researcher’s toolbox, and one Liu has been studying for more than 20 years.

The latest feature selection effort Liu is working on is led by doctoral student Jundong Li. The research has resulted in a widely used open-source repository, scikit-feature, and an influential article on feature selection algorithms and data sets. Li's work has been featured on leading machine-learning website KDnuggets as a must-see project. 

“Regardless of how powerful machine-learning algorithms are, we still need to get the data ready for learning,” Liu said.

For example, Liu’s former doctoral student Reza Zafarani, now an assistant professor at Syracuse University, collects minimal user information from multiple accounts to determine if accounts on different social media platforms are controlled by the same person. This practice can help identify malicious actors online.

Social computing with machine learning can also help to find the sweet spot between utility and privacy online without requiring users to compromise — something that doctoral candidate Ghazaleh Beigi is working on.

Currently, sharing more data online increases privacy and security risks. If we limit what we share, we would miss out on valuable opportunities to obtain targeted services, for instance, which can alleviate information overload. Beigi’s research demonstrates that a noncompromising approach to user privacy and utility is achievable.

Since 2007, Liu has also been involved in social computing research, an interdisciplinary field that is a natural extension of his research in data mining. By studying the intersection of user behavior on technical systems, researchers can gain new insights and provide new capabilities to design computer systems that support humans’ social behavior and interactions.

Liu’s students have leveraged social computing research to address a variety of challenges of life online.

Former doctoral student Pritam Gundecha, now with IBM Almaden Research Labs, conducts work that estimates whether a user online is vulnerable to privacy breaches. Doctoral candidate Liang Wu focuses on misinformation diffusion and detection.

Current graduate students Kai Shu, Lu Cheng and Kaize Ding have their own lines of research. Cheng aims to detect cyberbullying on social media. Shu studies automated fake-news detection on social media guided by social science theories in collaboration with ASU Professor H. Russell Bernard. Ding investigates outlier detection in social media analysis. 

Members of the Data Mining and Machine Learning lab pose for a group photo.

Students at all levels are producing valuable research in Huan Liu's Data Mining and Machine Learning Lab. Huan Liu poses with some of the many doctoral and graduate students on his lab team. Pictured: (seated) Isaac Jones, (from left) Raha Moraffah, Kai Shu, Ruocheng Guo, Professor Huan Liu, Lu Cheng, Tahora Nazer, Matthew Davis, Ghazaleh Beigi, Liang Wu, Nur Shazwani Kamarudin and Kaize Ding. Photo by Erika Gronek/ASU

Fostering academic leaders

Liu’s students’ work has gained recognition at ASU and beyond.

Two of his students have earned Dean’s Dissertation Awards, including Suhang Wang in fall 2018. Wang, now an assistant professor at Penn State, earned the award for his work exploring social media network representation learning.

In 2014, TweetTracker, a web-based system that collects and visualizes social media data for humanitarian assistance and disaster relief, earned the ASU President’s Team Award for Innovation. The project originated with two former doctoral students: Shamanth Kumar, who now works at Twitter, and Fred Morstatter, who works at the University of Southern California Information Sciences Institute.

In the wider research communities for social computing and feature selection, Liu’s students often are leaders in their field before they’ve even earned their doctorate; many have citation stats that rival those of some established faculty members. A high number of citations is one way to indicate their work is influential in their research community.  

“My senior doctoral students are actually postdocs in the lab,” Liu said. “They guide junior members and help them to succeed, and in the process, they naturally become leaders through their hard work. They can take on challenges to go straight to faculty positions, research labs and coveted companies. I’m proud of them, and they do ASU proud.”

Through the work with his doctoral students, Liu has one U.S. patent and has filed more than two dozen patent disclosures.

The Data Mining and Machine Learning Lab also hosts many undergraduate students who are working on their honors theses and research projects through the Fulton Undergraduate Research Initiative. That would not be possible without senior members of the lab taking on leadership roles in guiding and helping undergraduates. Liu says doctoral candidate Tahora Nazer is one of the latest outstanding strong research mentors among many former and current members of the lab team.

Gaining recognition in his community

The body of work Liu developed with these exceptional students recently earned recognition from three prominent professional organizations.

In the last few months of 2018, Liu was named a Fellow of the Association for Computing Machinery (ACM), the Association for the Advancement of Artificial Intelligence (AAAI) and the American Association for the Advancement of Science (AAAS). He is already a Fellow of the Institute of Electrical and Electronic Engineers (IEEE).

Election as a Fellow recognizes the researcher’s significant and influential contributions to important topics in the bestowing organization’s field.

“It’s a very pleasant surprise to get all three in the same year,” Liu said. “To get this kind of recognition is a reinforcement of our pursuit of excellence. It will help us reach a wider audience and also encourages us to be more creative and diligent.”

Within the Fulton Schools, Liu is on the faculty of the School of Computing, Informatics, and Decision Systems Engineering. He says the school’s leaders have always encouraged him to aim high in the goals for his research.

Liu also credits the environment of the School of Computing, Informatics, and Decision Systems Engineering, the Fulton Schools and ASU, which is enabling him to teach, collaborate and innovate with inspiring colleagues and excellent students who are motivated by high aspirations.

Monique Clement

Communications specialist, Ira A. Fulton Schools of Engineering


Hop to it: Researchers evaluate rabbits’ evolved resistance to myxoma virus

February 14, 2019

It's common knowledge that rabbit populations are not easily controlled — they reproduce swiftly, and as a result, they have a severe impact on their environment, as when European settlers introduced the wild European rabbit to Australia in the late 19th century. In an attempt to reduce the population size that had grown to almost a billion rabbits by 1950, Australian scientists released the myxoma virus — a virus known to be deadly to rabbits at the time — to the rabbit population, and eventually did the same for populations in France and the U.K. However, after some time, fatality rates lessened in all three countries, and the rabbit populations rebounded but were now genetically more resistant to the virus.

Regarded as “one of the greatest natural experiments in evolution,” researchers naturally wanted to learn more, so they tackled the genetic basis of the newly resistant rabbit adaptation to this virus. Grant McFadden's lab belongs to the Center of for Immunotherapy, Vaccines and Virotherapy. Download Full Image

Partnering with the University of Cambridge and several other research institutes, researchers at the Biodesign Institute at Arizona State University, as part of Grant McFadden’s Center for Immunotherapy, Vaccines and Virotherapy, validated the role of specific rabbit genes in contributing to this acquired resistance in research published in Science magazine.

McFadden’s lab has many decades of expertise in the myxoma virus, studying subjects ranging from the virus’s replication in hosts to its potential use in treating cancer. For this project, they were tasked with determining whether certain rabbit genes that had changed in the 70 years of exposure to the virus were responsible for the rabbits’ acquired resistance to the virus.

“There are rabbits in each population that evolved at the same time but independently of each other,” said McFadden, a professor in the School of Life Sciences and director of the Center for Immunotherapy, Vaccines and Virotherapy. “The idea was to sequence examples of many rabbit genomes of all three places and see what they have in common, and that’s what led to this study. We came up with half a dozen gene variations in common — our job was to determine whether these variants of genes affected that virus in a lab setting.”

While the rabbits that were resistant to the virus survived and thus were selected for, the less pathogenic viruses were also selected for among the viral populations. This coupled with the fact that the same trend was seen in three geographically distinct regions of the world, serve as a concrete example of co-evolutionary forces that operate between viruses and their hosts, and being able to determine the genetic basis of this adaptation only furthers our knowledge of parallel adaptation.

“The host and the virus began to do a genetic dance with each other that was started over 70 years ago. For decades after that, no one knew what that genetic dance was, but now we have learned something new from the genomes of the surviving rabbits,” McFadden added.   

The researchers in the U.K. were largely responsible for utilizing modern sequencing technology to sequence the rabbit genomes in the populations now and compare them to genomes from past generations, while McFadden and his lab were responsible for determining whether the genes that emerged in all three rabbit populations were correlated to antiviral effects by testing the virus in cell culture. Ana Lemos de Matos and Masmudur Rahman, a postdoc and an associate research professor in McFadden’s lab, respectively, were responsible for testing the effect of these genes on the myxoma virus. 

By doing so, the researchers were able to validate the role of these genes in viral replication and indicated that selection for a more effective interferon response as part of the innate immune response to viral infection in rabbit populations was at play.

McFadden and his lab believe that one of the main takeaways from this study was to prove that co-evolution happens and can occur quickly after new virus-host interactions emerge.

“This is probably one of the best examples of co-evolution that we know of, where the virus is evolving, and the host is evolving, and they are evolving in concert with each other,” McFadden said. “This is a wonderful example of pure-curiosity research, and there may be implications down the line, but in terms of co-evolution, I can’t think of a better example on the planet.”

Gabrielle Hirneise

Assistant science writer , Biodesign Institute