Computer Science Seminars

Hardware Acceleration for Realtime Robotics

In order for robots to achieve their full potential to assist humans, they need to be able to reliably operate in dynamic and unstructured real-world environments. Opportunities exist for roboticists to leverage hardware acceleration to overcome many of the computational bottlenecks that confine robots to the lab. In this talk I will show how Graphics Processing Units (GPUs) can be used to improve the performance of robotics algorithms through a case study of rigid body dynamics, and preview the potential for further acceleration through the use of custom Application Specific Integrated Circuits (ASICs).

Brian Plancher is a Ph.D. Candidate studying Robotics at Harvard University working with Vijay Janapa Reddi and Scott Kuindersma. His research is focused on developing and implementing open-source algorithms for dynamic motion planning and control of robots by exploiting both the mathematical structure of algorithms and the design of computational platforms (e.g., CPUs, GPUs, FPGAs). As such, his research sits at the intersection of Robotics and both Computer Architecture and Numerical Optimization. Brian also wants to improve the accessibility of STEM education and co-chairs TinyMLedu. He enjoys teaching and designing new interdisciplinary, project-based, open-access courses that lower the barrier to entry of cutting edge topics like robotics and TinyML. Brian enjoys spending his free time with his wife, daughter, and puppy, and ski racing in the winters.

Empowering People with Disabilities through Accessible User Interfaces

Computer science education has become widely popular as a gateway to higher education and lucrative careers. In learning computer science, students can develop life skills like problem-solving, critical thinking, and logical thinking. Computer science education can also empower people with disabilities to gain technical knowledge to solve accessibility barriers that they may face in the future, enhance their communication skills, and lead to more self-determination and independence.

This talk will provide an overview of accessibility and Human-Computer Interaction research exploring multimodal tools for inclusive learning and supporting independent living. The projects discussed in this talk will describe creating novel research prototypes that utilize computer vision and augmented reality and analyze them through the lens of three key perspectives in Human-Computer Interaction: user-centered design, participatory design, and ability-based design.  The talk will conclude with a discussion on future research areas around designing shareable displays for multi-user contexts and designing educational technologies to expand awareness around privacy and security. This research can lead to accessible adaptive assistive technologies for people with disabilities in the future. 

Varsha is a Ph.D. Candidate in Computer science at the University of Colorado Boulder. She is advised by Shaun Kane, and her research interests lie at the intersection of Human-Computer interaction and accessibility. Varsha’s research focuses on creating accessible smart technologies that empower people with disabilities in their everyday activities. Currently, she is exploring how Augmented Reality(AR) can support people with cognitive disabilities to customize supportive prompts in their daily routines. Varsha’s research has been published at premier international HCI conferences such as CHI, ASSETS, IDC, and VL/HCC, including a Best Paper Honorable Mention Award at CHI 2019. Her work has been recognized by the University of Colorado Boulder through the Hope Schutz Jozsa Award, the PhD Summer Research Fellowship, and the Carol Lynch Memorial Fellowship. She has also been a Coleman Fellow at the Coleman Institute for Cognitive Disabilities. Varsha is also the computer science department’s lead teaching assistant (2021-22). Previously, she received her M.S in Computer Science from CU Boulder.

Considering Process in Algorithmic Bias Detection and Mitigation 

Image

Artificial Intelligence (AI) systems now affect important decisions in people's lives, from the news articles they read, to whether or not they receive a loan. While the use of AI may lead to great accuracy and efficiency in the making of important decisions, recent news and research reports have shown that AI models can act unfairly: from exhibiting gender bias in hiring models, to racial bias in recidivism prediction systems. 

In this talk, I’ll discuss methods for understanding fairness issues in AI through considering the process by which models arrive at their decisions. This technique contrasts with a large portion of AI fairness literature, which focuses on studying model outcomes alone. Specifically, I will show how considering a model’s end-to-end decision process allows us to expand our understanding of unfair behavior---such as in my work demonstrating how model instability can lead to unfairness by having important decisions rely on arbitrary modeling choices (e.g. whether or not a person is granted a loan from a decision-making model may depend on whether some unrelated person happened to be in the training set). Secondly, I will discuss how considering process can help us find bias mitigation techniques which avoid a tradeoff between predictive utility and fairness, with case studies from my collaborations with Stanford RegLab and the Internal Revenue Service (IRS) investigating tax auditing practices, and with Cornell, Microsoft Research, Upturn, and others to examine the role of criminal risk assessment models in racial disparities in pre-trial detention.

Emily Black is a PhD candidate in the Accountable Systems Lab at Carnegie Mellon University, advised by Matt Fredrikson. Her research centers around understanding the impacts of machine learning and deep learning models in society. In particular, she focuses on showing ways in which commonly used machine learning models may act unfairly; finding ways to pinpoint when models are behaving in a harmful manner in practice; developing ways to mitigate harmful behavior when possible; and translating technical insights into technology policy recommendations. She is currently supported by an Amazon Graduate Research Fellowship. (For more information, please see https://www.cs.cmu.edu/~emilybla/).

Limits and Applications of Natural Language Processing

Natural Language Processing (NLP) is a branch of Artificial Intelligence that aims to build machines that humans can seamlessly interact with through spoken and written language. As NLP becomes more ubiquitous in our daily lives, through technologies like Google Translate and Apple’s Siri, understanding the limits of these systems is critical. In the first half of the talk, we will cover my work developing diagnostics that test such systems. With the remaining time, we will discuss how my research group has applied Natural Language Processing to leverage data from social media (e.g. Twitter and Reddit) in fields like psychology, public health, and nursing. 

Adam Poliak completed his Ph.D. in Computer Science at Johns Hopkins University where he was a member of the Center for Language and Speech Processing. He is currently a Roman Family Faculty Fellow at Barnard College, Columbia University in the Computer Science Program where he teaches courses in Data Science and runs a research group focused on Natural Language Processing. He has done research stints at the MIT Lincoln Laboratory,  Bloomberg L.P.,  and the Human Language Technology Center of Excellence. Adam's research has been funded by the FDA, Google, Columbia School of Nursing, and DARPA, and others. Prior to graduate school, Adam spent time at numerous start-ups.

Polynomial Counting in Anonymous Dynamic Networks with Applications to Anonymous Dynamic Algebraic Computations

In recent years, the problem of Counting the number of nodes in Anonymous Dynamic Networks (ADNs) has attracted a lot of attention. ADNs are algorithmically challenging because nodes are indistinguishable (they lack identifiers) and the topology is adversarial (i.e. network links may change arbitrarily from a communication round to the next one) limited only to maintain the network connected. Counting is central in distributed computing because the number of participants is frequently needed for algorithmic decisions, such as termination, agreement, and synchronization.

A variety of distributed algorithms built on top of mass-distribution techniques have been presented, analyzed, and also experimentally evaluated; some of them assumed additional knowledge of network characteristics, such as bounded degree or given upper bound on the network size. However, the question of whether Counting can be solved deterministically in sub-exponential time remained open until recently.

In this talk, I will present our recent Methodical Counting Algorithm for ADNs, which runs in polynomial time and requires no knowledge of network characteristics. I will also show how to extend Methodical Counting to compute the sum of input values and more complex functions without extra cost. The Methodical Counting Algorithm and its extensions to other algebraic and Boolean function computations were the first that could be implemented in practice on large ADNs with worst-case guarantees. I will also overview some of the successful follow-up work on this and other lines of research involving undergraduate students.

Miguel A. Mosteiro is an Associate Professor of Computer Science Pace University. Before, he was Assistant Professor at Kean University, Research Professor at Rutgers University and Research Fellow at  the Universidad Rey Juan Carlos (Spain). He obtained his PhD in Computer Science from Rutgers University, and his BE in Electronics from Universidad Tecnologica Nacional (Argentina). His research interests span the broad areas of algorithms and distributed computing. He focuses on algorithms for restricted wireless networks, crowd computing, cloud computing, and more recently spammability and fairness of ranking algorithms such as PageRank. His research work is conducted in collaboration with scholars from various countries, involving graduate as well as undergraduate students. Prof. Mosteiro has received the 2016 Kean University Undergraduate Research Mentor of the Year Award, concurrently with the Undergraduate Researcher of the Year Award for one of his students.

Knowledge-enhanced Text Generation: The Curious Case of Figurative Language and Argumentation 

Large-scale language models based on transformer architectures, such as GPT-3 or BERT, have advanced the state of the art in Natural Language Understanding and Generation. However, even though these models have shown impressive performance for a variety of tasks, they often struggle to model implicit and/or non-compositional meaning, such as figurative language and argumentative text. In this talk, I will present some of our recent work on text generation models for figurative language and argumentation. There are two main challenges we have to address to make progress in this space: 1) the need to model common sense and/or connotative knowledge required for these tasks; and 2) the lack of large training datasets. I will discuss our proposed theoretically-grounded knowledge-enhanced text generation models for figurative language such as metaphor and simile, as well as for enthymeme reconstruction and if time permits argument reframing. I will conclude by discussing opportunities and remaining challenges for incorporating knowledge in neural text generation systems.

Smaranda Muresan is a Research Scientist at the Data Science Institute at Columbia University and an Amazon Scholar. Before joining Columbia, she was a faculty member in the School of Communication and Information at Rutgers University where she co-founded the Laboratory for the Study of Applied Language Technologies and Society. At Rutgers, she was the recipient of the Distinguished Achievements in Research Award. Her research interests are in computational semantics and discourse, particularly figurative language understanding and generation, argument mining and generation, and fact-checking. Most recently, she has been interested in applying NLP to education and public health, as well as in building NLP technologies for low resource languages. She received best papers awards at SIGDIAL 2017 and ACL 2018 (short paper). She is currently serving as a board member of the North American Chapter of the Association for Computational Linguistics (NAACL) and as a Program Co-Chair for ACL 2022.

Building the Audio Programmer: Power through Inexperience, Transparency, and Sharing

In 2017, Joshua Hodge started a YouTube channel called The Audio programmer to teach a skill that he knew hardly anything about - audio software development.  Since then, The Audio Programmer has become a central hub for audio developers of all levels. How is this possible?

In this talk, Joshua Hodge will discuss lessons learned while creating the Audio Programmer - how his inexperience became the best experience, how being excluded allowed him to be inclusive, and the ultimate power of sharing your work.  

Algebraic Problems: The Frontier of Efficient Randomized Computation

Randomness is a valuable resource in many computational tasks. Indeed, the security and/or the accuracy of many randomized algorithms and protocols rely on the random bits being truly random and independent. However, in practice such random bits are elusive, which may compromise the performance of the underlying systems. This motivates the following fundamental question:-

Can every computational task that requires randomness be carried out deterministically, paying, perhaps only a small overhead?

Meanwhile, the nature of many algebraic problems makes them amenable to randomized algorithms. For example: a random set of vectors is independent, a random assignment to a low-degree polynomial is non-zero etc. Thus, you can easily find a set of independent vectors and a non-zero assignment by picking them uniformly at random. Indeed, it is not surprising that the frontier of efficient randomized computation consists of algebraic problems. Among the frontier problems are Polynomial Identity Testing, Polynomial Factorization and others.

In this talk, I will discuss my research on the relationship between randomness, computation and algebra. Time permitting, I will also discuss the problems I have been working on and some recent connections to cryptography and machine learning.

Dr. Ilya Volkovich is a Senior Lecturer in the Department of Computer Science and Engineering at the University of Michigan, where he has taught courses in the theory of computation for several years. Previously, he was a Postdoctoral Research Associate in the Computer Science Department at Princeton University and held a visiting position at the Institute of Advanced Study. In 2012, he obtained his Ph.D. in Computer Science from Technion, Israel Institute of Technology, advised by Prof. Amir Shpilka. His research interests are in the broad area of theoretical computer science and discrete mathematics. More specifically, he is interested in aspects of algebraic complexity, randomness in computation, computational learning theory, and their applications to cryptography and machine learning.