Department of Computer Science and Statistics

Undergraduate Research

---

A Web-based Toolkit for Exploring Cryptography
Dr. Ed Lamagna

Cryptography is an important, contemporary real-world application of computer science and mathematics.  The subject can be taught to a wide range of student audiences.  To assist in teaching and learning, a set of web-based tools that perform encryption and decryption with a variety of classical and contemporary methods has been developed.  The tools were used successfully in both a freshman honors class for non-majors and an upper level class for computer science majors.  The site provides a powerful and uniform environment for exploring the topics commonly taught in introductory courses on cryptography.  These include substitution and transposition ciphers, block codes (a simplified version of DES), public key infrastructure and techniques (Diffie-Hellman, RSA), and hashing.  There are also tools for exploring underlying number theoretic concepts such as modular arithmetic and the Euclidean algorithm.  The site provides tools both to trace visually, step-by-step, how the methods operate, and computational tools to perform cryptanalytic attacks on classical ciphers and small instances of public-key ciphers.  An instructor can use the tools in the classroom to explain the algorithms and to present examples.  Students use the site to explore the methods on their own, to solve problems, and to crack cryptographic challenges.  The tools eliminate the need for students to write programs to perform these computational tasks, enabling them to focus on important algorithmic and mathematical ideas.

---

Accelerating Approximate Search on Large, High-Dimensional Datasets
Dr. Noah Daniels

A novel method of accelerating approximate search on large, high-dimensional datasets, demonstrating applications in both bioinformatics and astronomy. This approach builds on mathematical and computational foundations including fractal geometry, hierarchical clustering, and data compression. This approach is called CHESS (Clustered Hierarchical Entropy-Scaling Search). This research resulted in a paper with two undergrads as co-authors at IEEE Big Data 2019.

---

Manifold Mapping
Dr. Noah Daniels

Manifold mapping builds on the work described above in a way that enables anomaly and outlier detection. The manifold mapping algorithm is called CLAM (Clustered Learning of Approximate Manifolds) and it enables a collection of ensemble-learning anomaly-detection algorithms called CHAODA (Clustered Hierarchical Anomaly and Outlier Detection Algorithms), of course pronounced “chowda.” CHAODA outperforms state of the art methods on 14 of 18 datasets under testing.  This research, done with an undergrad, has been submitted as a paper to ICML, which is currently under review

---

Extension of CHESS
Dr. Noah Daniels

This project is a major extension of the CHESS paper, involving a rewrite in Rust, many new distance functions (including a fast implementation of n-dimensional Wasserstein distance), efficient data compression, and extension to many public datasets in biology and astronomy, including image data from the Sloan Digital Sky Survey’s MANGA project. This project is in collaboration with a colleague at the University of Navarra, Spain. The new tool is called CLAM-CAKES (because Rhode Island, after all).  

---

Recognize
Dr. Krishna Venkatasubramanian

This project involves developing an app called Recognize that teaches individuals with intellectual disabilities about various types of abuse, how to recognize it and what to do to mitigate it. The project has three undergraduates (working with a senior developer) who are part of the development team who are utilizing concepts from HCI and software engineering to successfully manage a complex software development project. The team is getting ready to perform user studies on this app with actual individuals with intellectual disabilities. 

---

Individuals with Upper Extremity Mobility Impairment Navigating Mobile Devices During COVID
Dr. Krishna Venkatasubramanian

This project involves understanding how individuals with upper extremity mobility impairment are navigating their use of mobile computing technologies during the COVID pandemic. For instance how does working from home, PPEs, masks etc have affected their technology use. The project has one undergraduate student (working with a PhD student) who is helping with interviewing actual people with upper extremity mobility impairment to understand their lived experience with respect to computing technology use during this pandemic. This work is being prepared to be submitted to the ACM ASSETS conference in April.

---

Using Alexa to Train Individuals with Intellectual Disabilities to Make 911 Calls
Dr. Krishna Venkatasubramanian

This project involves two undergraduate students investigating the use of Amazon Alexa to be able to train individuals with intellectual disabilities to make 911 calls. The idea is for Alexa to act as the 911 operator and ask questions. This goes beyond asking questions from a static list. The challenge here is to make Alexa ask questions in an adaptive fashion that takes in account the response of the person with intellectual disabilities. This project is in the exploratory stage where we are trying to understand Alexa’s ability to be adaptive.

---

Understanding How People Use Smart Home-Safety Devices
Dr. Krishna Venkatasubramanian

Smart home-safety devices such as locks, cameras etc. are being heavily used by people with disabilities to make their lives easier. We want to understand how people use these smart home-safety devices. The project has one undergraduate student (working with a PhD student) who is looking at Amazon reviews (over 40,000 reviews are being analyzed) for several classes of smart home-safety devices and using qualitative analysis methods to understand the larger trends of the experiences of people with disabilities in using these smart home devices. This work is being prepared to be submitted to the OzCHI conference in August.

---

Detecting overlapping patterns in Asteroid, a programming language which supports both first-class and conditional pattern matching
Dr. Lutz Hamel

Asteroid is a multi-paradigm programming language which supports first-class patterns. As such Asteroid supports functional-style pattern matching in multi-dispatch functions.  This capability allows for pattern overlap to occur which can lead to incorrect program behavior.  This project is to detect pattern overlap and help the user to improve their code.  This work has been accepted to the flagship conference of the Council on Undergraduate Research: Posters on the Hill.   Nationwide only sixty participants were selected.

---

Self Organizing Maps and Locally Linear Embedding for Dimensionality Reduction
Dr. Lutz Hamel

This was an undergraduate research project to see how our implementation of unsupervised learning with neural networks compares to other industry standard algorithms, in particular, locally linear embedding.  The study used a particularly tricky data set called the “swiss roll” data set.  The results were presented at the Sigma Xi Undergraduate Research Conference in 2015 and  won best poster in the category of computer science, mathematics, and engineering.

---

Convergence Analysis of VSOM: A High-Performance, Stochastic Training Algorithm for Self-Organizing Maps
Dr. Lutz Hamel

This was a systematic study of the convergence behavior of VSOM, a high-performance, stochastic training algorithm for self-organizing maps.  The study was conducted by an undergraduate and led to changes in the algorithm resulting in convergence behavior similar to the industry standard training algorithms available for self-organizing maps.  The changes have been published as part of our POPSOM package and are now available as part of R.

---

Current Undergraduate Researchers