Prospective Student Visit Day
The Department of Computer Science thanks everyone who visited our annual Prospective Student Visit Day and Graduate Research Symposium on Friday, November 13! It was wonderful to see so many skilled people eager to learn about our MCS and PhD programs.
Students who were interested in our graduate programs met with our faculty and listened to presentations about the current research of our PhD candidates and professors. These topics ranged from cryptography to healthcare and showcased first-hand the broad applications of the University of Iowa’s research community.
This year’s events included the 12th Annual Computer Science Graduate Research Symposium, which brought together five speakers to present their research to a wide audience of computer science students, graduates, and faculty.
Graduate Research Symposium
Below are the researchers who presented at the Graduate Research Symposium, alongside the titles of their projects and the abstracts.
| Presentations | Gallery |
|---|---|
Robert Lorch, PhD candidate. “Leveraging Deductive Verification for Cryptographic Protocols.” Abstract: Cryptographic protocols are hard to prove correct. Defining the notion of protocol correctness is itself nontrivial, even before attempting a proof. The general difficulty of the problem has resulted in flawed “proofs” of correctness, sometimes leading to real-world security impacts. Even worse, security concerns are often eschewed in favor of financial considerations, and so proofs are often not attempted in the first place. In an ideal situation, one would like to define a security protocol, propose a security property, and have some automated tool either prove the protocol correct or return a concrete attack trace, all with minimal human interaction. While great progress has been made, current tools and techniques are still inadequate to achieve this goal. In particular, the approaches with the highest levels of automation are the most difficult to interact with in cases of verification failure, and approaches that offer usable interactivity in the case of verification failure are the least automated. In this talk, we discuss an attempt to achieve the best of both worlds—that is, to provide high automation potential while still offering a promising path forward when (automated) verification fails. Advisor: Cesare Tinelli | Areas of research: Automated Reasoning, Computer Security. |
|
Hieu Vu, PhD candidate. “Generating and Transforming Temporal Graphs for Healthcare Modeling.” Abstract: Modern hospitals generate patient clinical records and operational data about patient locations, healthcare workers, rooms, treatments, and potential interactions among them over time. These data can support patient-outcome prediction, infection tracking, operational planning, and intervention evaluation. Yet high-quality operational data are difficult to collect and share because they are sensitive, costly to obtain, and governed by strict privacy requirements. Even when data are available, their scale and changing structure make machine learning computationally expensive and meaningful patterns difficult to extract. My research addresses these challenges through temporal graphs, which represent entities and their changing relationships. In this talk, I will introduce two complementary approaches: generating realistic synthetic hospital mobility data and transforming large dynamic graphs to make training downstream models more efficient. I will show how incorporating hospital-specific information improves the usefulness of synthetic data for disease-spread simulation, and how graph coarsening can reduce computational cost while maintaining predictive performance. Together, these approaches work toward making complex hospital data more accessible and practical for healthcare research. Advisor: Bijaya Adhikari | Areas of research: Data-centric machine learning on temporal graphs for healthcare. |
|
Ian Pope, PhD candidate. “End-to-End EEG-Based Auditory Attention Decoding for Cochlear Implant Users." Abstract: Cochlear implant (CI) users continue to experience difficulty understanding speech in complex listening environments. Auditory attention decoding (AAD) from EEG offers a promising pathway toward neurofeedback-based training, but remains challenging in CI users due to limited data, implant-related artifacts, and high inter-subject variability. In this work, we investigate end-to-end convolutional neural network (CNN) models for decoding auditory attention directly from single-trial EEG, without access to the auditory stimulus. We introduce a sequence of task-aligned, EEGNet-based architectural refinements, culminating in a multi-scale, one-dimensional temporal–spatial model tailored to auditory attention dynamics. Across a cohort of 105 CI users, the proposed architectures substantially outperform traditional temporal response function (TRF) baselines, achieving average improvements of 17.30% in accuracy and 28.10% in AUC. Strategies targeting label noise, data scarcity, and inter-subject and inter-trial variability provide complementary gains, but architectural design is the primary driver of performance. These results demonstrate that carefully designed EEG-only CNN architectures can reliably decode auditory attention in CI users, supporting practical neurofeedback systems for real-world listening scenarios. Advisor: Octav Chipara | Areas of research: Intelligent Hearing Technology |
|
Sarmad Chandio, PhD candidate. “Role of Online Platforms in Shaping Public Opinion.” Abstract: Social media platforms have become an integral part of our lives, influencing how individuals consume news, discover products, and interact with peers. Since the process is so seamless, we have essentially offloaded parts of our cognitive and decision-making processes to platform algorithms. This increased reliance on systems has enabled them to shape our opinions and ideologies. However, these “powerful systems” largely operate as opaque black boxes, leaving users and researchers with little visibility into how recommendations are generated and how they drive shifts in human behavior. This talk investigates the black-box algorithms powering major online platforms and examines their impact on user opinions and ideologies. I leverage language as a behavioral proxy, analyzing what users watch, produce, and consume to understand ideological shifts. Advisor: Rishab Nithyanand | Areas of research: Computational Social Science - Platforms, Ideologies, Audits |
|
Weiran Wang, assistant professor. “Tuning Distributionally Robust Learning via Bilevel Minimax Optimization.” Abstract: Deep learning models increasingly rely on Distributionally Robust Optimization (DRO) to combat subpopulation shifts and spurious correlations. However, advanced DRO methods depend on sensitive hyperparameters that traditionally demand computationally costly grid searches. We propose a framework that treats the robustness mechanism itself as a learnable component, continuously tuning these parameters on held-out validation data directly within the training loop. Because robust training is inherently a minimax game, optimizing its hyperparameters creates a complex nested bilevel minimax structure. To solve this, we introduce a novel penalty-based first-order method that operates efficiently without relying on restrictive lower-level strong convexity assumptions. Furthermore, we provide a generalization guarantee for our method, which dynamically discovers the bias-variance trade-off on the fly, bypassing the discretization errors and massive computational overhead of grid searches. This talk covers joint work with Yiyang Shen (PhD@Informatics), Qihang Lin (Tippie College of Business), and AI. Bio: Weiran Wang is an Assistant Professor in the Department of Computer Science (with joint affiliations in AMCS and Informatics) at the University of Iowa. Prior to joining academia, he spent seven years in industrial research positions. He received his PhD from the University of California, Merced and was a postdoctoral researcher at the Toyota Technological Institute at Chicago (TTIC). His research takes an algorithmic approach to machine learning, focusing on multimodal representation learning, generative models, optimization for machine learning, and their applications to speech processing and large-scale language models. |
|
These speakers offered prospective students a glimpse into the world of graduate research and how our department is innovating in a variety of fields.
We are proud of our current PhD candidates for their well-made presentations and we're excited to see how their research progresses! We are also grateful to all the undergraduates, prospective graduate students, and faculty who participated in the event and made it such a success!