Click the links that follow for more news about Nishanth Kumar, previous recipients of honorable mentions for this award, and other recent accomplishments by our students.
The Computing Research Association (CRA) is a coalition of more than 200 organizations with the mission of enhancing innovation by joining with industry, government, and academia to strengthen research and advance education in computing. Every year, they recognize North American students who show phenomenal research potential with their Outstanding Undergraduate Researcher Award, and in 2020, Brown CS made one of the strongest showings in the Honorable Mentions category. Out of 103 students who received Honorable Mentions, four of them are Brown CS students: Deniz Bayazit, Rigel Galgana, Nishanth Kumar, and Esteban Safranchik. Last year, Brown CS students received three Honorable Mentions.
Asked to desribe her work, Deniz says that it lies in the intersection of Natural Language Processing (NLP) and robotics. "Humans," she says, "can convey instructions to other humans via different means of communication. They have the ability to enter a new environment and follow through a navigation task when given a partial or full map, and hinted with a form of natural language (NL). Reproducing a similar ability in a robot would be extremely useful for untrained users, who do not have an in-depth knowledge of robot programming. To achieve grounding language to arbitrary landmarks in any outdoor environment (or in other words, without needing to re-train the language model for a new environment), I worked with Matthew Berg, Rebecca Mathew, Ariel Rotter-Aboyoun, Professor Pavlick, and Professor Tellex in summer, 2019. We are currently investigating spatial language understanding models that could help guide agents in partially observable environments with natural language hints such as 'The car is behind the CIT on Waterman Street.'"
"Let's play a game," says Rigel when posed with questions about his research. "Say that you are in dire need of TA assistance on a project, so you decide to register for every single TA hours slot for the next month in hopes of getting seen at least once. While you would ideally want a high spot on SignMeUp so as to avoid a long wait, the queue is shuffled randomly at the beginning of each set of hours. After this shuffling, you see how long you'll have to wait before getting seen and decide whether or not to wait it out or delete your ticket. Assuming you knew in advance how crowded each session was ahead of time, how do you decide when to wait it out? How long do you expect to wait? What if you wanted to get seen at least twice? Thrice? This variant of the secretary problem is only one of many applications of order statistics. Other applications include auction design and analysis, Bayesian-optimal pricing, and more. For the last year and a half, I've been working with Professor Amy Greenwald on efficient algorithms to compute joint distributions of order statistics with the goal of answering some of the above questions."
Nishanth's research has focused on two interrelated components of AI for Robotics: intelligence and collaboration. "On the collaboration front," he says, "I've worked on using Augmented and Virtual Reality (AR/VR) to allow humans to effectively and intuitively give robots all the information required to accomplish complex tasks. On the intelligence front, I've helped make state-of-the-art Learning from Demonstration (LfD) algorithms more data-efficient. I've also worked on bridging the fields of Reinforcement Learning and Formal Methods to allow agents to reason about which parts of the world are relevant to their decision making, and thus ignore irrelevant factors."
"I'm incredibly lucky and grateful," says Nishanth, "to be part of such an amazing community within Brown CS: none of my work would be possible without it. Specifically, I want to thank my amazing mentors and collaborators in the H2R, IRL, and RLAB groups, especially Professors Stefanie Tellex, George Konidaris, and Michael Littman. I also want to thank TStaff and CIS for never failing to support the unique needs that come with running experiments on robots. I couldn't imagine a better environment in which to produce great research!"
"My primary area of research," Esteban tells us, "is weakly supervised machine learning, a paradigm for machine learning featuring noisy or limited information sources as substitutes to full supervision. My work with Prof. Stephen Bach addresses the challenges of building classifiers and structured predictors for NLP in the absence of hand-labeled data. During the summer, I worked on WISER, a framework for training deep sequence taggers with weak supervision from user-written rules. Our work introduces a novel approach to modeling weak supervision sources that achieves state-of-the-art performance on several NLP tasks. WISER allows users to build powerful machine learning systems without relying on ground-truth labeled data."
The full list of Outstanding Undergraduate Researcher winners is available here.
For more information, click the link that follows to contact Brown CS Communication Outreach Specialist Jesse C. Polhemus.