Samsung and CDS Research Day

Researchers from the Samsung and CDS meet to share research and plan future collaborations

Spring is traditionally known as the season of rebirth, rejuvenation, and growth. Trees shrug off the long winter and stretch their arms towards the blue sky, flowers bloom, and birds are all a-flutter with joy — in every place but New York, it seems! Spring has been slow to grace the Big Apple this year, but as we wait for her to arrive, a spring-like atmosphere of new research activity, collaborations, and partnerships is flourishing at CDS. Just last Friday, our faculty met with employees from the Samsung Research Center to share recent developments in the field and organize future research collaborations between the two institutions.

CDS Director Richard Bonneau kicked off the day with an introduction about CDS, followed by introductions about the Samsung Research and AI Center, which was launched in 2017 after combining the organization’s Software R&D Center and Digital Media & Communications Center. Key areas of focus for Samsung’s new research hub include the Internet of Things (IOT) and artificial intelligence. These research themes are, of course, also near and dear to CDS faculty and students, as was demonstrated through several topic presentations for Samsung employees after the introductions.

For example, Joan Bruna, Assistant Professor of Mathematics and Data Science (and recent recipient of the prestigious Sloan Foundation Research Fellowship!), presented on his work with graph neural networks, a powerful technique which that has been used to make significant advancements in areas like particle physics, quantum chemistry, and recommendation systems.

Bruna was followed by Sam Bowman, Assistant Professor of Data Science and Linguistics, who explained his research on using latent tree learning to teach machines to understand sentences to a meaningful depth — a crucial feature for those who are hoping to invent personal assistant robots like Siri and Alexa, or to computationally perform translations, summarizations, sentiment analysis. Similarly, Kyunghyun Cho, Assistant Professor of Computer Science and Data Science (and CIFAR Azrieli Global Scholar), demonstrated how he has been using neural networks to improve multilingual machine translations.

After a hearty lunch, student researchers who have been working under the guidance of CDS professors provided detailed explanations about how they have applied new approaches and methodologies to specific projects. Bruna’s graduate student researcher, Nicholas Choma, for example, revealed how he has been using graph neural networks for IceCube, a neutrino observatory whose goal is to detect high-energy neutrinos (they are specific subatomic particles) that originate from black holes.

Bowman’s graduate student researcher, Nikita Nangia, showed how she created ListOps, a diagnostic dataset for latent tree learning, while Cho’s graduate student researchers Jason Lee and Elman Mansimov highlighted their work on non-autoregressive sequence models for machine translations for texts, as well as audio and image captioning (their model, by the way, boasts a 90% translation rate!)

As a whole, the day clearly showed that Samsung and CDS have much to gain from each other. New research collaborations between us are hovering on the horizon — and we here at CDS could not be more excited to see what the future holds.

by Cherrie Kwok