Allen School News

After a pandemic-enforced hiatus, last week the Allen School welcomed industry partners, alumni and friends to its 2022 Research Showcase this week to celebrate the groundbreaking work of its students and faculty. The typically annual event, which is hosted by the Industry Affiliates Program, welcomes industry partners and alumni to engage with the school’s research and learn more about how its members are advancing the field of computing. This year marked the first time the showcase has been held since 2019. 

The agenda included a variety of technical sessions featuring the latest and greatest Allen School research. Topics ranged from computing for the environment and artificial intelligence for health, to robotics and computational biology. The day concluded with an open house and poster session that culminated in the announcement of the Madrona Prize and People’s Choice Awards. 

Allen School professor Yejin Choi, who was recently named a MacArthur Fellow for her work in natural language processing and commonsense AI, gave the keynote address titled “David vs. Goliath: The Art of Leaderboarding in the Era of Extreme Scale Neural Models.” During the talk, she highlighted the power of smaller neural models developed in academia and how they can have an edge over larger industry-scale models.

One of the approaches that gives them this edge, Choi explained, is Symbolic Knowledge Distillation, a new framework she and her colleagues proposed that distills knowledge symbolically as text besides just the neural model alone. Her work produced a machine-authored commonsense model that, for the first time, surpassed a human-authored model in all criteria, including scale, accuracy and diversity. Instead of humans directing the commonsense knowledge graph, Choi and her collaborators found that through their framework machines could write their own knowledge graph, teaching themselves to distill language, knowledge and reasoning. 

Choi also explained how an unsupervised, inference-time reasoning algorithm, can match or surpass supervised approaches on hard-reasoning tasks or complex language generation tasks that require logical constraints. The algorithm, called NeuroLogic Decoding, illustrated the limits of larger-scale neural models while also demonstrating better performance on text-generation tasks. 

While bullish on the potential of smaller, high-quality models, Choi acknowledged that scaling is a necessary condition for making progress in AI — necessary, but not sufficient.

“Scaling laws are real, and denial is futile,” said Choi, the Bret Helsel Professor in the Allen School and senior research manager at the Allen Institute for AI. “Especially when we think about hard problems in AI, we cannot just solve the hardest problems by scaling things up — analogous to how you cannot reach to the moon by making the tallest building in the world one inch taller at a time.”

In the evening, nearly 300 people came together in the Paul G. Allen Center to catch up after two years of pandemic-enforced absence and view the latest research from Allen School labs. More than 50 teams of student researchers shared their work with attendees, who were invited to vote for their favorite poster or demo as part of the People’s Choice Award. 

Members of Madrona Venture Group, longtime friends and supporters of the Allen School, were on hand to present the Madrona Prize, which recognizes exciting projects with commercialization potential. Madrona partner Chris Picardo announced that members of the  UW Database Group behind VOCAL — short for Video Organization and Interactive Compositional AnaLytics — captured top honors for two projects related to that work. VOCAL allows users to extract semantic content from large datasets while minimizing inefficiencies in data cleaning, compositional queries, exploration and organization.

“We’re delighted to be back and able to award the prize again and have an amazing poster session,” Picardo said. “We’re thrilled that we get to do this and get to see such amazing research.” 

2022 Madrona Prize

Winner

Video Organization and Exploration and Interactive Video Analytics for Compositional Queries: Ph.D. students Maureen Daum, Enhao Zhang and Dong He; Allen School director and professor Magdalena Balazinska; professor Ranjay Krishna; alum Brandon Haynes (Ph.D., ‘20), now a senior scientist at Microsoft

Runners up

Clearbuds: Wireless Binaural Earbuds for Learning-Based Speech Enhancement: Ph.D. students Ishan Chatterjee, Maruchi Kim and Vivek Jayaram; professors Shyam Gollakota, Ira Kemelmacher-Shlizerman, Shwetak Patel and Steven M. Seitz

Data Analysis Systems for Statistical Non-Experts: Ph.D. student Eunice Jun; professors Jeffrey Heer and René Just

Levity: Contactless Robotics and Automation for Synthetic Biology: Ph.D. student Jared Nakahara; professor Joshua R. Smith

2022 People’s Choice Awards

Distributing Trust and Establishing Transparency in Societal Scale Computing and Network Infrastructure: Ph.D. students Sudheesh Singanamalla, Matthew Johnson and Esther Han Beol Jang; master’s students Nick Durand and Abhishek Shah; postdoctoral researcher Spencer Sevilla; professors Richard Anderson and Kurtis Heimerl

Beyond WEIRDness of NLP: Ph.D. student Sebastin Santy; professors Katharina Reinecke and Yulia Tsvetkov; alum and former AI2 postdoc Maarten Sap (Ph.D., ‘21); psychology professor Andrew Meltzoff; research scientist Ronan Le Bras of the Allen Institute for AI; undergraduate alum Jenny Liang (B.S., ‘21), now a Ph.D. student at Carnegie Mellon University; research scientist Rodolfo Barragan of the UW Institute for Learning & Brain Sciences

Read GeekWire’s coverage of the awards here.

Thank you to our friends at Madrona and all of the members of the extended Allen School community who joined us in celebrating and supporting student innovation — it was wonderful to see you again! Read more →

When Allen School professor Simon Shaolei Du opened his inbox on a cloudy Friday morning in October, he wasn’t expecting anything out of the ordinary, let alone a trip around the world. 

The website had slated the announcement for late September. When the date came and went, Du put the thought from his mind. Yet there it was. A click, a glow, then a smile. 

“It was very short notice, actually,” he said, grinning again. 

Next, plane tickets and slide decks awaited him – along with, if he was lucky, he said, some barbecue. 

Du was recently named a 2022 Samsung AI Researcher of the Year for his contributions to the field of artificial intelligence. He is one of five recipients of the award, which recognizes promising researchers under 35 who have made a significant impact in advancing the discipline. 

On Tuesday, Du was formally recognized at the 2022 Samsung AI Forum held in Seoul, South Korea. His talk focused on results from his previous work with deep learning and reinforcement learning. 

“I am grateful for receiving this award as a researcher working on the theoretical foundations of artificial intelligence,” Du said. “Furthermore, I am thrilled that machine learning theory research is valued.”

His work was among the first to show why over-parameterized neural networks can be optimized by simple algorithms, such as gradient descent, and established a theory for explaining why deep learning works well in terms of generalization. While over-parameterized models, which possess more parameters than training examples, have shown their power in recent years, it is unclear why. Du’s research centers on answering this question. 

“I aim to develop unifying theories for over-parameterized models to identify their benefits and drawbacks,” he said.

Du added he and his collaborators have also explored reinforcement learning theory and are seeking to further their scholarship in this arena. Reinforcement learning mimics the way humans learn – through trial and error. A system of positive reinforcers (awards) and negative reinforcers (punishments) guide the neural network in adapting to stimuli within the environment. As the agent experiences more, it grows wiser. 

It’s a solution whose beauty lies in the details. 

“I love the elegance of math, and I would like to make a real-world impact,” said Du, who credited his collaborators and students for helping make this achievement possible. “Machine learning is a field where mathematics is crucial in designing practically relevant methods.”

Du joined the Allen School faculty in 2020. In addition to the Samsung AI Researcher of the Year Award, he has received a National Science Foundation CAREER Award, an AAAI New Faculty Highlights Award from the Association for the Advancement of Artificial Intelligence and a NVIDIA Pioneer Award, among others. 

Congratulations, Simon! Read more →

For first-generation college students, navigating the complexities of higher education can be intimidating. Often there isn’t a blueprint from which to work. Yet for the more than 30% of undergraduates across the University of Washington who are first-generation students, it’s far from a solo journey. 

In honor of the National First-Generation College Celebration taking place today, the Allen School is highlighting members of its community who are among the first in their families to pursue a college degree. Here are a few of their stories. 

Responses have been edited for length and clarity.

Kent Zeng, undergraduate student

Kent Zeng, a senior studying computer science and minoring in math, serves as co-chair of GEN1, an organization for first-gen students in the Allen School. The life he’s living now contrasts sharply with the one his parents left behind when they immigrated to the U.S. from China. It’s a fact that spurs him, he says, to give back to his family and his community.

Allen School: Please describe why being a first-generation student is meaningful to you.

Kent Zeng: Being immigrants, my parents never really pursued their passions and mainly focused on providing for me and my sister in hopes that we could build a career that we found fulfilling. Being first-generation is meaningful to me because I am getting opportunities that my parents could only wish for. I honestly feel bad sometimes since I am getting all these experiences in college and my parents never did. However, I think the real point I should be focusing on is using these experiences to give back to my parents and to my community.

Allen School: How did you become involved with GEN1?

KZ: I initially got involved in GEN1 because I wanted to meet new people with a similar background as me. Now as an upperclassman my reason for staying involved in GEN1 is to support other first generation Allen School students academically and professionally. I’ve just started my role as co-chair, but the GEN1 team is kicking off the year strong. I recently led the Drive Your Career with Uber Technical Workshop where first-gen students learned from Uber recruiters and Uber technologists about how to best apply to technical roles and what day-to-day life at Uber is like. This year GEN1 is also hosting monthly community socials for our members to connect and we’ve got a lot planned for National First Gen Day! 

Allen School: Can you speak to the future impact of being a first-generation student?

KZ: I think one of the main generational outcomes that comes from being a first-generation student is the fact that the trajectory of families can dramatically change for the better. For starters, while not always the case, having a college degree can allow students from low-income families to graduate and work in lucrative industries, fundamentally altering the way families live. Education also expands minds, and first-generation students are now often able to think in ways that previous generations could not even fathom. In addition, with education, first-generation students can then go on to raise the foundation on which future generations can continue to grow.

Allen School: What is your favorite part about being a student at the Allen School? 

KZ: My favorite part of being a student in the Allen School is being connected with individuals who have similar motivations as I do and who challenge me to do better. Just hearing about all the cool things people in the Allen School community are doing inspires me to also do better and use my skills to help others.

Sonia Fereidooni, B.S./M.S. student

Sonia Fereidooni (B.S., ‘22) is a B.S./M.S. student studying computer science and one of the co-founders of GEN1. In June, she graduated with bachelor’s degrees in computer science and sociology. Now working on her master’s at the Allen School, she has served as a teaching assistant and was a lecturer for CSE 373: Data Structures and Algorithms this summer. 

Her love of teaching comes naturally; Fereidooni hails from a family of teachers. Her mother taught mechanical engineering and her grandmother taught high school math in Iran. Being the first in her family to receive a degree in the U.S. is not something she takes lightly. “It amplifies a story that anyone can make it,” she says, “and the more we are vocal about being the first in our family to achieve a postsecondary degree in the U.S., the more other first-generation students are inspired to do the same.”

Allen School: What inspired you to pursue a university degree?

Sonia Fereidooni: What inspired me was witnessing a female Iranian mathematician’s journey when I was a child. I first learned about Dr. Maryam Mirzakhani when I was about 10 or 11 years old, through an email chain my mother sent me about inspiring Iranian women.

Throughout the next few years, the more I learned about Dr. Mirzakhani, the more I wanted to pursue the field of mathematics like her and one day become a research professor. In 2014, she became the first woman to win the Fields Medal, the highest honor in mathematics, which led me to think that nothing was impossible. 

Allen School: What advice would you give first-generation students about to begin their college journey?

SF: I would advise that first-generation students understand that they are not like most of their peers entering the Allen School and this can be an empowering trait. Most first-generation students come into the Allen School already knowing how to build with intent and for the betterment of their larger community. Giving back to an underprivileged community is what many first-generation students have in mind and have as a goal when thinking about success at the Allen School. And there is strength in having that mindset as part of a large community of first-generation students, because we will always strive to support one another. 

Allen School: What is your favorite part about being a student at the Allen School?

SF: I have seen the Allen School in many different lights, but my favorite part is how fluid the school can be with leveraging opportunities to students. Even though the Allen School may seem very structured and traditional since it is one of the largest, most prestigious and most established computer science programs in the world, it really is not bound to some set of rules and structure the way most university departments are. At times, you can just reach out to professors and advisers and have a chat as friends. Or there will be opportunities where you obtain a research position just by talking to a research professor in a 20-second elevator ride at the Allen Center. The school is large, but if you are able to leverage the network and opportunities, you will be successful. 

Sofia Isadora Padilla Muñoz, PMP student

Sofia Isadora Padilla Muñoz is a student in the Allen School’s Professional Master’s Program. Hailing from Guadalajara and raised in San Julian, both cities in Mexico, she earned her bachelor’s degree in electronics engineering from the Tecnológico de Monterrey. She cherishes her education, she says, because she knows it’s not a given. “Being a first-generation student is meaningful to me because I broke the endless family cycle where women, like my mother, were not allowed to study,” she says. “I am very grateful for my parents because they made it possible that my sisters and I are empowered and independent women, where each one can decide about her life and own it to live a fulfilled one.”

Allen School: What inspired you to continue your education after earning your bachelor’s degree?

Sofia Isadora Padilla Muñoz: I wanted to explore other areas of computer science to see which one I would like to specialize in further. Also, I wanted to connect with other engineers and hear their histories. Before coming to the Allen School, I was working as a software developer at Microsoft.

Allen School: What initially sparked your interest in computer science, and why did you choose the Allen School in particular? 

SPM: I decided to study electronic engineering and computer science because I thought that solving problems to improve the quality of life of people through engineering would add meaning and fullness to my life. So far, I think that it does. That is why I considered to keep studying and specializing in order to have a greater impact and responsibility in the new problems that society faces. I wanted an in-person program and, in verifying the rankings and academic consultants, the UW was the best option for me. It has all that I require to keep studying and working at the same time.

Allen School: What advice would you give first-generation students about to begin their college journey?

SPM: I would recommend to first-generation students to go to psychological counseling because we need help processing the changes and understanding them as a difficult challenge and not as a threat. Additionally, I would recommend they learn the best they can instead of being the best. In the end, what you learn is what you will remember. I have reached a point in my life where I no longer remember all the awards and prizes I have received because they simply do not define me. Knowledge and emotional stability give you security; prizes and awards do not. Finally, I would also add to have fun and enjoy the journey of college life and life itself: make friends, feel love, laugh loud, cry as needed, run, rest and eat. We only have one body and one mind and we need to take care of it.

Elle Brown, staff

In the days of dial-up, Elle Brown, a graduate advising program coordinator in the Allen School, remembers learning how to program, reinstalling Windows, tinkering with that heavy box, their portal to another world. It was not expected that they go to college, Brown says, but they had a passion for knowledge and a perseverance that allowed them to earn their bachelor’s degree after dropping out of high school. Now at the Allen School, Brown is helping others navigate their respective academic journeys. 

Allen School: Please describe why being a first-generation student is meaningful to you.

Elle Brown:  Being a first generation student is not only part of my history but part of my ancestors’ as well. My father did not graduate from high school. The story he told was that he was standing in the graduation line to walk in the ceremony when the administration pulled him out and said he was missing one credit. This infuriated him, so he walked out and never went back. Shortly after, he joined the army. My mom was a bride at 14. Her first husband, not my father, was 30 years her senior. She wouldn’t meet my father until she was 22. Needless to say, she also did not graduate high school.

Allen School: What are some challenges you experienced? 

EB: I dropped out of high school when I was 16 due to a number of factors, including chronic depression and anxiety. I was living with my paternal grandmother at the time, and she was still overcome with grief over my grandfather’s death five years prior. I was too much for her to deal with, and it was easier to let me drop out of school than to fight me every day, according to her. I managed to get a job at McDonald’s making $6.10 an hour, in hopes of moving out on my own and to stop being a burden on my grandmother. Though I lived in a rural area where the cost of living was considerably less than Seattle, I was still not able to make ends meet. I earned my GED in 2001, and I tried to figure out how to pursue college. I wasn’t sure how else to get on my own feet. I opted to apply to DeVry University in the school’s information technology subfields. A recruiter even came out to my grandma’s house to talk to us about enrolling.

Allen School: You overcame many obstacles on your way to graduating with your bachelor’s degree from UW. Can you talk a bit more about your journey from Georgia to earning your degree in Washington? 

EB: As a high school dropout and GED recipient, it was difficult to find colleges in my area that would accept my application. DeVry, though, seemed more than happy to take me on. My grandma and I didn’t understand the for-profit aspect of DeVry, and I know now that getting into that program was not about qualifying as much as it was about my being able to pay. My family was not able to provide any financial support. They never told me I would go to college, or that I needed to work hard to get there. It was simply not a possibility. The only people who talked about college were in books, those in TV or movies and at the occasional presentation at public school. There were student loans, though, and I could be declared independent since I didn’t have parents in my life. I would take on the loan debt. At the time there was a program called the HOPE Scholarship in Georgia that began in 1993. It was funded by the state lottery and promised money to students who showed academic excellence. Unfortunately, as a GED-recipient I only qualified for a tiny stipend.

I did not graduate from DeVry, and in fact I didn’t make it through a whole quarter. It was a rather dark time. In 2003, I met someone online and he asked me to move to Oregon to live with him. It was after I arrived there that I applied to Portland Community College and tried again to earn a degree. I did end up graduating from PCC in 2013 with an associate’s degree in general studies. I applied to UW in 2016 to earn a bachelor’s degree. I almost didn’t make it through my bachelor’s, and if it wasn’t for my adviser I would have dropped out my final quarter.

Allen School: What experiences led you to a position with the Allen School?

EB: My academic adviser was helping me search for job opportunities. I had the pleasure of interning with her during the last quarter of my degree in 2019. She found the program coordinator job with the Allen School and encouraged me to apply. I met with Elise, Garrette, who was the program coordinator before me, Les, Jen Hiigli and an outside staff member. They liked me and here I am!

Joe Eckert, staff

Joe Eckert, the Ph.D. program manager in the Allen School, spent about eight years waiting for the federal government to acknowledge his financial independence while he attended community college as a part-time student. Once he was able to take loans out under his name, he finished his undergraduate work at Humboldt State University, now Cal Poly Humboldt. 

He continued his education at UW, pursuing a doctorate in geography. His experience as a graduate student eventually led him to academic advising. “After spending so much time and energy to uncover the ‘hidden curriculum’ for myself, I decided to continue teaching it to others as a staff adviser,” he says, noting his shift from student to staff. “I had to muck through this journey on my own and I’m excited to be able to help others avoid that.”

Allen School: What inspired you to pursue higher education?

Joe Eckert: Telemarketing is not a “forever” job, no matter how good you are at it.

Allen School: What led you to a position with the Allen School? 

JE: While I was supposed to be writing a dissertation, I picked up a graduate assistantship working as an academic adviser to geography undergraduates. I quickly learned that I liked advising far more than writing dissertations and took a year away from the program to take a temporary advising job at the Undergraduate Academic Affairs advising office, working primarily with non-majors who wanted to pursue tech careers. When the position concluded, I took on a role at the iSchool advising future librarians in their master’s program. In the middle of the pandemic, this job became available and I applied as quickly as I could!

Allen School: What is your favorite part about being a member of the Allen School? 

JE: I enjoy helping students learn that “hidden curriculum” that was invisible to me when I started both undergrad and graduate work. I’m privileged to work in a position that allows me to challenge “that’s just the way it’s always been” at both an interpersonal as well as a structural level. Working with folks who are discovering this academic way of being is the best. 

Allen School: What advice would you give first-generation students about to begin their college journey?

JE: Talk to someone. Find someone in your life who has been to college. It may not be your parents and it may not be anyone in your extended family. But you don’t have to navigate applications and career-planning alone – even if you’re used to having to figure stuff out independently. Also college advisers are nothing like high school counselors. You should ask your adviser things!

Learn more about UW’s first-generation celebration here. Read more →

If you’re a frequent flyer, you may have amassed a motley collection of complimentary airline earbuds over the course of your travels — if you didn’t toss them in the trash immediately after clearing customs, that is. Soon, those throwaway pieces of plastic and wire could potentially transform the lives of children around the world. A team that includes Allen School professor Shyam Gollakota and Ph.D. student Justin Chan has devised a way to repurpose inexpensive earbuds to turn any smartphone into a device for screening newborn babies for hearing loss. The team described its prototype system in a paper published today in Nature Biomedical Engineering.

Clinicians screen newborns’ hearing by stimulating and evaluating otoacoustic emissions, or OAE, which are the sounds generated by the movement of the outer hair cells of a healthy cochlea. While such screening is currently routine in the United States, in many countries hospitals and clinics cannot afford the specialized equipment for administering the test. That disparity, along with firsthand experience, motivated Gollakota to tackle the project.

“I grew up in a country where there was no hearing screening available, in part because the screening device itself is pretty expensive,” Gollakota, who holds the Washington Research Foundation / Thomas J. Cable Professorship in the Allen School, told UW News. “The project here is to leverage the ubiquity of mobile devices people across the world already have — smartphones and $2 to $3 earbuds — to make newborn hearing screening something that’s accessible to all without sacrificing quality.”

The team designed an inexpensive probe using the aforementioned earbuds, an off-the-shelf microphone and a length of lightweight silicon tubing. The system plays a different tone through each earbud to stimulate the cochlea, and then records the OAE via the attached microphone and transmits it through the smartphone’s headphone jack for processing. In clinical testing involving more than 100 patients, the UW-developed system performed as well as commercial equipment costing thousands of dollars.

Now that they have a prototype, the researchers’ next step is to partner with local experts to scale up their approach. The team has already made a start by partnering with the UW Department of Global Health, the University of Nairobi and the Kenya Ministry of Health on the TUNE project, short for Toward Universal Newborn and Early Childhood Hearing Screening in Kenya. As reported on TUNE’s website, in countries with routine newborn screening programs, doctors detect hearing impairments in babies by the time they are two to three months of age; in countries without such programs, hearing loss often is not detected until the child reaches three years of age.

“We have an opportunity to really have an impact on global health, especially for newborn hearing,” Chan said. “I think it’s pretty gratifying to know that the research we do can help to directly solve real problems.”

Chan and Gollakota’s co-authors include Dr. Randall Bly and Dr. Emily Gallagher, both affiliated with UW Medicine and Seattle Children’s; Dr. Nada Ali of UW Medicine; Ali Najafi, who earned his Ph.D. from the UW Department of Electrical & Computer Engineering; Anna Meehan of Seattle Children’s, and Lisa Mancl of the UW Department of Speech & Hearing Sciences.

Read the Nature Biomedical Engineering paper here, a related article here, and the UW News release here. Read more →

When COVID consigned doctor-patient interactions from the clinic to a computer screen, Allen School Ph.D. candidate Xin Liu already had his finger on the pulse of that paradigm shift. Since his arrival at the University of Washington in 2018, Liu has worked with professor Shwetak Patel in the UbiComp Lab to combine mobile sensing and machine learning to real-world problems in health care, with a focus on developing non-contact, camera-based physiological screening and monitoring solutions that are accessible by all and adaptable to wide range of settings. His goal is to “democratize camera-based non-contact health sensing for everyone around the world by making it accessible, equitable, and useful.”

Liu embarked on his undergraduate education in the US as an international student realizing how difficult it was to integrate into a new culture. As an undergraduate student at UMass Amherst, he became the first international student consultant and peer-mentor. He also encouraged other international students to embrace leadership roles. Liu recounted that, “this experience motivated my research in computer science and health where I aimed to develop useful and accessible computing technologies for diverse populations.”

Liu’s research would take on new meaning and urgency as the pandemic upended modern social interactions. As remote clinical visits increased significantly during this time, the need for remote ways of sensing and monitoring heart health became critical for medical practitioners and patients alike. Liu’s work combining camera-based physiological sensing with machine learning algorithms could offer new possibilities in early detection of heart-related health issues as well as allow for much-needed remote diagnostics when a patient faces barriers to obtaining in-person care at a clinic — even outside of a pandemic. 

As Liu saw it, there were several major issues that needed to be considered, and, in some cases, corrected in order for camera-based health assessment to be widely applicable. From a practical standpoint, the tool would have to obtain a high level of accuracy for medical practitioners to evaluate patients’ vital signs remotely in order to make informed clinical decisions. Privacy is another major critical concern when it comes to people’s personally identifiable medical information; because of this, any collected data would need to be held locally on the device.

That device could come with varying capabilities — or lack thereof. 

“Since people have access to a wide range of devices, the application has to function on even the most rudimentary smartphone,” Liu explained. “Likewise, people in resource-constrained settings may not consistently have connectivity, so the application would need to be capable of running without being connected to a network.”

Disparate access to resources is not the only consideration for Liu and his colleagues when it comes to equity.

“In the past, camera-based solutions were skewed towards lighter skin tones, and did not function well with darker skin tones,” Liu noted. “To be truly useful, particularly to populations who are already underserved in health care, our application has to function accurately across the full range of skin tones, and under a variety of conditions.”

In 2020, Liu and his fellow researchers proposed the first on-device neural network for non-contact physiological sensing. This novel multi-task temporal shift convolutional attention network (MTTS-CAN) addressed the challenges of privacy, portability and precision in contactless cardiopulmonary measurement. Their paper, which was among the top 1% of submissions to the 34^th conference on Neural Information Processing Systems (NeurIPS), was foundational in that it allowed for health sensing on devices with lower processing power. Following this, Liu conceived an even faster neural architecture called EfficientPhys for lower-end mobile devices, which will appear at the 2023 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). 

Last year, Liu proposed an unsupervised few-shot learning algorithm called MetaPhys, presented at the Conference on Health, Inference, and Learning (CHIL), and a mobile-camera based few-shot learning mobile system called MobilePhys, which appeared in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT), as steps toward addressing remote sensing’s shortcomings with regard to variations in patients’ skin tones, activities and environmental contexts. He has been involved with Microsoft Research’s efforts to use synthetic avatars to simulate facial blood flow changes and to systematically generate a large-scale video-based dataset under a wide variety of conditions such as different skin tones and backgrounds. This work has led to the production of a synthetic data set that offers labels with precise synchronization and without noise to overcome issues with variability and diversity. 

Having proven the concept, Liu has turned his attention to ensuring such tools will be reliable and efficient for diverse populations under real-world conditions. Whereas previous research on non-contact camera-based physiological monitoring has focused on healthy populations, Liu has initiated a collaboration with Dr. Eugene Yang, clinical professor and director of the Eastside Specialty Clinic at UW Medicine, to collect data in in-patient and out-patient clinical settings. Their goal is to validate the team’s machine learning augmented, camera-based approach for obtaining accurate readings for a range of health indicators, such as heart rate and respiration rate, in real-world clinical settings. Ultimately, Liu aims to push the boundaries of non-contact physiological measurement through exploring contactless camera sensing to measure such readings as blood pressure and arrhythmias.

“Xin takes a collaborative and interdisciplinary approach to his research,” said Patel, Liu’s Ph.D. advisor. “He works closely with his clinical partners to inform his research and executes his research with the highest ethical and equitable standards. He has taken on some difficult research challenges around skin tone diversity for health-related AI models that are already having industry impact.” 

Liu earned a 2022 Google Fellowship in Health Research and Artificial Intelligence for Social Good and Algorithmic Fairness to advance this work and foster the completion of his dissertation work. 

Congratulations Xin! Read more →

Growing up in Japan, Akari Asai never imagined that she would one day pursue a Ph.D. at the Allen School focused on developing the next generation of natural language processing tools. Asai hadn’t taken a single computing class before her arrival at the University of Tokyo, where she enrolled in economics and business courses; her first foray into computer science would come thousands of miles from home, while studying abroad at the University of California, Berkeley. The experience would alter the trajectory of her academic career and put her on a path to solving problems on a global scale.

“I changed my major in the middle of my undergraduate studies, and I wished I had discovered computer science and opportunities for pursuing my career abroad earlier,” said Asai. “My own situation made me realize the importance of information access for everyone.”

That realization led Asai to pursue her Ph.D. at the University of Washington, where she is now in the business of developing next-generation AI algorithms that offer rich natural language comprehension using multi-lingual, multi-hop and interpretable reasoning working with Allen School professor Hannaneh Hajishirzi in the H2Lab.

“Akari is very insightful and cares deeply about the impact of her work,” observed Hajishirzi, who is also senior research manager in the Allen Institute for AI’s AllenNLP group. “She is bridging the gap between research and real-world applications by making NLP models more efficient, more effective, and more inclusive by extending their benefits to languages other than English that have been largely ignored.”

More than 7,100 languages are spoken in the world today. While English is the most prevalent, spoken by nearly 1.5 billion people, the global population is nearing 8 billion — meaning a significant proportion is excluded from the benefits of today’s powerful NLP models. Asai is trying to close this gap by enabling universal question answering systems that can read and retrieve information across multiple languages. For example, she and her collaborators introduced XOR-TyDi QA, the first large-scale annotated dataset capable of open-ended information retrieval across seven different languages other than English. The approach — XOR QA stands for Cross-lingual Open Retrieval Question Answering — enables questions written in one language to be answered using content expressed in another. 

Asai also contributed to CORA, the first unified multilingual retriever-generator framework that can answer questions across many languages — even in the absence of language-specific annotated data or knowledge sources. CORA, short for Cross-lingual Open-Retrieval Answer Generation, employs a dense passage retrieval algorithm to pull information from Wikipedia entries, irrespective of language boundaries; the system relies on a multilingual autoregressive generation model to answer questions in the target language without the need for translations. The team incorporated an iterative training method that automatically extends the annotated data previously only available in high-resource languages to low-resource ones. 

“We demonstrated that CORA is capable of answering questions across 28 typologically different languages, achieving state-of-the-art results on 26 of them,” Asai explained. “Those results include languages that are more distant from English and for which there is limited training data, such as Hebrew and Malay.”

Language is not the only barrier Asai is working to overcome. The massive computational resources required to operate the latest, greatest language models, which few groups can afford, also puts them out of reach for many. Asai is making strides on this problem, too, recently unveiling a new multi-task learning paradigm for tuning large-scale language models that is modular, interpretable and parameter-efficient. In a preprint, Asai and her collaborators explained how ATTEMPT, or Attentional Mixture of Prompt Tuning, meets or exceeds the performance of full fine-tuning approaches while updating less than one percent of the parameters required by those other methods.

Asai is also keenly interested in the development of neuro-symbolic algorithms that are imbued with the ability to deal with complex questions. One example is PathRetriever, a graph-based recurrent retrieval method that learns to retrieve reasoning paths over the Wikipedia graph to answer multi-hop open-domain questions at web scale. By leveraging a reading comprehension model alongside the retriever model, Asai and her colleagues enabled PathRetriever to explore more accurate reasoning paths in answer to complex questions compared to other methods. Some of her co-authors subsequently adapted the system to enable complex queries of scientific publications related to COVID-19. 

Ultimately, Asai intends to integrate the various facets of her research into a general-purpose, lightweight retriever and neuro-symbolic generator that will be capable of performing complex reasoning over diverse inputs while overcoming data scarcity. Having earned a 2022 IBM Ph.D. Fellowship earlier this year to advance this work, Asai’s ambition is to eliminate the disparity between the information “haves” and “have nots” by providing tools that will empower anyone to quickly and easily find what they need online — in multiple languages as well as multiple domains.

“Despite rapid progress in NLP, there are still several major limitations that prevent too many people from enjoying the benefits of that progress,” she explained. “My long-term research goal is to develop AI agents that can interact with broad swaths of internet users to answer their questions, giving everyone equal access to information that might otherwise be limited to certain default audiences.”

Her commitment to promoting equal access extends beyond information retrieval to include the field of NLP itself; to that end, Asai is an enthusiastic mentor to students from underrepresented backgrounds.

“I’m excited to continue making progress on my own research interests,” said Asai, “but I hope to also inspire the next generation of researchers in AI.”

Way to go, Akari! Read more →

Allen School professor Jeffrey Heer received a Test of Time Award at the 2022 IEEE Visualization & Visual Analytics Conference this week, marking the third consecutive year that his work has been recognized with the honor. 

Heer, the co-director of the Interactive Data Lab, co-authored the winning paper, “Enterprise Data Analysis and Visualization: An Interview Study,” which provided key insights into understanding how data analysts operate and the challenges they encounter in their workflows. The paper was published in IEEE Transactions on Visualization and Computer Graphics in 2012 when Heer and two of his co-authors were researchers at Stanford University. The team presented its findings at the IEEE Conference on Visual Analytics Science and Technology (VAST).

The Test of Time Award recognizes papers published at previous conferences that have had a lasting impact both within and beyond the field of visualization over the ensuing decade. 

“We’re thrilled that our colleagues still find our work relevant 10 years later,” said Heer, who holds the Jerre D. Noe Endowed Professorship at the Allen School. “While research ‘style’ probably plays a part, I think we were also in the right time and place as societal interest in data and analysis heated up. I also think it helps that our work has had interdisciplinary teams spanning human-computer interaction, data management, cognitive science and other disciplines, bringing more varied perspectives.” 

As the team members investigated tools to aid data analysis, they were particularly interested in what Heer referred to as “the messy early stages of data cleaning and preparation.” While they had hands-on experience themselves, they wanted to gain a deeper understanding of the different types of analysts, the issues they encountered and the infrastructure that supported them. 

To tackle this problem, the team launched a qualitative study, interviewing participants across a range of sectors, including health care, retail, social networking, finance, media, marketing and insurance. Among the insights they uncovered included analyst archetypes, how analysts collaborated with one another and with other business units and challenges they met in the analysis process. 

“The study helped prioritize our thinking around the varied backgrounds of people working with data, including different levels of statistical and software development expertise, and how organizational dynamics shapes how the work gets done,” Heer said. “Both are important to consider if new tools are going to be usable and effective in practice.”

Heer credited first author Sean Kandel, his former Ph.D. student at Stanford and fellow co-founder of Trifacta, with leading the project. Kandel was the chief technical officer of Trifacta, which develops data wrangling software and a popular data preparation platform used by Google, NASA and others. Trifacta was acquired by Alteryx in February. 

Co-author Andreas Paepcke is the director of data analytics and senior research scientist at Stanford. Co-author Joseph M. Hellerstein is the Jim Gray Professor of Computer Science at UC Berkeley and a co-founder of Trifacta. Hellerstein was at UC Berkeley at the time of the study. 

Heer previously earned Test of Time Awards at VIS for his work on data-driven documents and narrative visualization. Since joining the Allen School faculty in 2013, he has been recognized with the Association for Computing Machinery’s ACM Grace Murray Hopper Award, the IEEE Visualization Technical Achievement Award and Best Paper Awards at the ACM Conference on Human Factors in Computing (CHI), EuroVis and IEEE InfoVis conferences. Last year, Heer was elected to the CHI Academy by the ACM Special Interest Group on Computer-Human Interaction (SIGCHI) for his substantial contributions in the areas of data visualization, data wrangling, text analysis, language translation and interactive machine learning. 

Read Heer’s latest award-winning paper here and the award citation here. 

Congratulations to Jeff and his co-authors! Read more →

Science News has named professor Huijia (Rachel) Lin, a founding member of the Allen School’s Cryptography group, as one of its SN 10: Scientists to Watch. Each year, Science News recognizes 10 scientists who are making a mark in their respective fields while working to solve some of the world’s biggest problems. Lin earned her place on the 2022 list for achieving a breakthrough on what has been alternately referred to as the “holy grail” or “crown jewel” of cryptography by proving the security of indistinguishability obfuscation.

“I’m very attracted to these magical concepts,” Lin told Science News. “The fun of it is to make this concept come to realization.”

While Lin explores a variety of fundamental problems — from black-box constructions for securing multiparty computation to zero-knowledge proofs — her work on iO has been celebrated for answering an open question that had vexed cryptographers for more than 40 years: How to prove the security of this potentially powerful “master tool” for securing data and computer programs while maintaining their functionality and bring it into the mainstream. According to the article, previous attempts at proving iO were generally geared toward obtaining a result that would be deemed “good enough” — and one by one, those attempts would unravel under further scrutiny. 

Lin aimed for more than “good enough” by seeking a generalizable solution grounded in sound mathematical theory. Rather than approaching iO like “a bowl of spaghetti,” as she put it, Lin preferred to attack the problem by untangling it into its component parts, working alongside University of California, Los Angeles professor Amit Sahai and his then-Ph.D. student and NTT Research intern Aayush Jain, now a professor at Carnegie Mellon University. After two years, the team had a theoretical framework for provably secure iO that was based on a quartet of well-founded assumptions: Symmetric External Diffie-Hellman (SXDH) on pairing groups, Learning with Errors (LWE), Learning Parity with Noise (LPN) over large fields, and a structured-seed Boolean Pseudo-Random Generator (sPRG). The result was first reported in Quanta Magazine in the summer of 2020; Lin and her collaborators subsequently earned a Best Paper Award at the Association for Computing Machinery’s 53rd Symposium on Theory of Computing (STOC 2021) for their contribution — one that Lin is eager to see progress from theory to reality.

“These are ambitious goals that will need the joint effort from the entire cryptography community,” she observed at the time. “I look forward to working on these questions and being part of the effort.”

In other words, watch this space.

Lin is the second member of the Allen School’s Theory of Computation group to be recognized on the SN 10 list, after her colleague Shayan Oveis Gharan was highlighted in 2016.

Read Lin’s profile in Science News here, and the Quanta article on the i/O breakthrough here.

Photo: Dennis Wise/University of Washington Read more →

The Allen School has established a new center at the University of Washington that aims to catalyze the next generation of cloud computing technology. The Center for the Future of Cloud Infrastructure, or FOCI, will cultivate stronger partnerships between academia and industry to enable cloud-based systems to reach new heights when it comes to security, reliability, performance, and sustainability.

“The first generation of the cloud disrupted conventional computing but focused on similar engineering abstractions, which is typical of many new technologies,” said Allen School professor Ratul Mahajan, co-director of the FOCI Center and, until recently, co-founder and CEO of cloud computing startup Intentionet. “Now that cloud computing is on the cusp of a more radical transformation, this center will help usher in a new era by cultivating tighter partnerships between researchers and practitioners to address emerging bottlenecks and explore new opportunities.”

That transformation is being driven in large part by the rise in machine learning, edge computing, 5G and other burgeoning technologies. According to Mahajan’s Allen School colleague and center co-director Simon Peter, the demands of these new workloads — including exponential growth in the energy required to power their applications — will require researchers to rethink the full computing stack from the ground up. 

“Companies and consumers are seeking ever-greater levels of security, reliability and performance in the cloud at a reduced cost,” Peter noted. “Not just monetary cost, but also in terms of cost to the environment. For a while, thanks to Moore’s Law, we were gaining ground when it comes to energy efficiency. But now the gains have slowed or even reversed; for example, in the U.S. the energy demand for computation is growing twice as fast as solar and wind power. So we need to think holistically about the hardware-software interface and how to make cloud computing sustainable as well as resilient and secure.”

One of the areas that Peter and his colleagues are keen to explore is energy-aware cloud computing, which would enable tradeoffs between power and performance while making cloud applications resilient to disruption. Another potential avenue of inquiry concerns how the development of systems to effectively manage the variety of hardware accelerators used in settings such as disaggregated storage and emerging machine learning applications while minimizing latency, ensuring fairness, and meeting multi-dimensional resource needs — among other challenges.

How the center approaches these challenges will be informed by a technical advisory board comprising representatives of cloud companies Alibaba, Cisco, Google, Microsoft and VMware — all significant movers and shakers in the cloud space. Their input will help guide the center’s research toward real-world impact based on current trends, what problems they anticipate over a five to 10- year time horizon, and how solutions might be applied in practice. Center researchers will apply these practical insights to their pursuit of big, open-ended ideas, drawing upon cross-campus expertise in systems, computer architecture, networking, machine learning, data science, security, and more.

“Industry knows the pain points and technology trends; academia is adept at the exploratory, collaborative work that’s fundamental to solving hard problems,” noted Allen School professor and center co-director Arvind Krishnamurthy, who also serves as an advisor to UW machine learning spinout OctoML. “By bringing the two together, this center will not only yield compelling solutions but also contribute to the education of students who will go on to build these next-generation systems.”

The FOCI Center was seeded with industry commitments totaling $3.75 million over three years. The Allen School is hosting a launch event on the UW campus in Seattle today to connect faculty and student researchers with industry leaders interested in shaping the future of cloud computing.

“Seattle is the cloud city, both in weather and as home to the largest cloud companies, so it was only natural to establish a center focused on cloud computing and leverage the synergies between the UW’s research expertise and our local industry leadership of this space,” said Magdalena Balazinska, professor and director of the Allen School. ”When it comes to what we can accomplish together, I would say the sky’s the limit.”

To learn more, visit the FOCI Center website and read the coverage by GeekWire here.

Main photo credit: University of Washington Read more →

Yejin Choi, a professor in the Allen School’s Natural Language Processing group, was selected as a 2022 MacArthur Fellow by the John D. and Catherine T. MacArthur Foundation to advance her work “using natural language processing to develop artificial intelligence systems that can understand language and make inferences about the world.” The MacArthur Fellowship — also known as the “genius grant” — celebrates and invests in talented and creative individuals whose past achievements signify their potential to make important future contributions. Each recipient receives a stipend of $800,000 that comes with no strings attached.

“It’s been several weeks since I learned about this award, and it still feels so surreal,” Choi told UW News.

Choi, who joined the Allen School faculty in 2014, may feel like she is dreaming, but her work has had a very real impact. Currently the Brett Helsel Career Development Professor in the Allen School and senior research manager for the Mosaic team at the Allen Institute for AI, Choi has contributed to a series of high-profile projects that have expanded the capabilities of natural language models — and uncovered potential pitfalls. For example, she was among the first to bridge the fields of NLP and computer vision by teaching models to generate original and accurate image descriptions based on visual content in place of conventional statistical approaches. She has also contributed to a variety of tools for analyzing and combating the proliferation of bias and misinformation online, from  AI-generated “fake news” to trashy training inputs that lead to toxic language degeneration, along with new methods for assessing the quality of open-ended machine-generated text compared to that generated by humans.

Choi and her collaborators went a step further with the development of Ask Delphi, an experimental platform for exploring how machines might acquire and exercise moral judgment in response to real-world situations. Through this and other work, Choi is pushing the field closer to her overarching goal: to imbue machines with a human-like ability to reason and communicate about the world in both physical and abstract terms. Whatever comes next, Choi is determined to fulfill the spirit of the Fellowship by pursuing the most original and impactful research ideas — even when they are accompanied by a degree of risk.

“Taking the road less traveled may seem exciting at first, but sustaining this path can be lonely, riddled with numerous roadblocks and disheartening at times,” Choi said. “This fellowship will power me up to go ahead and take that adventurous route.”

Previous MacArthur Fellowship winners with an Allen School connection include Choi’s faculty colleague Shwetak Patel, alumni Stefan Savage (Ph.D., ‘02), a professor at the University of California San Diego, and Christopher Ré (Ph.D., ‘09), a professor at Stanford University, and former Allen School faculty member Yoky Matsuoka, currently founder and CEO of Yohana.

Read the UW News release here and check out Choi’s MacArthur Foundation profile here. Read the New York Times story here, GeekWire article here and Crosscut interview with Choi here.

Congratulations, Yejin! Read more →