Gain invaluable hands-on research experience and take your skills to the next level. The Early Research Scholars Program (ERSP) offers a unique opportunity for aspiring researchers to work on groundbreaking projects alongside leading experts in their field.
The Early Research Scholars Program, initially designed by the University of California, San Diego (UCSD), is a group-based, team-mentored research experience for students early to midway through the major program. Students will be placed into small groups and then matched with an active research project from the CS department to give students the opportunity to observe the research process in action. Then, students will propose and complete a related research project that they will also present at the end of the program.
Over two semesters, gain in-depth knowledge of research fundamentals and observe an active research project under the guidance of an expert instructor, a central mentor, and a research mentor.
We promote and support a diverse Computer Science community along many dimensions, including but not limited to gender, race, ethnicity, nationality, socioeconomic situations, ability, sexual orientation, or challenges facing first-generation students attending college. Our goals is to help students overcome barriers such as financial obligations, lack of role models, and community support, among other hardship or challenges.
The ERSP program spans over two consecutive semesters, the fall term and then the spring term.
Introduces students to research in Computer Science and CS-related disciplines. Students will learn how to conduct literature searches, read research papers, and perform analyses. Students will work in teams to identify and formulate research problems, design and conduct research studies, analyze results, and present the research orally and in written form.
CSC 490 will count as an advanced major elective course toward your CS graduation requirements.
Provides students with the opportunity to complete the research project proposed in CSC490. Students will conduct research activities including collecting data, running experiments, getting results, and performing analyses. Students will present their research orally and in written form.
CSC491 can fulfill your CS Capstone requirement.
Joseph Delgado, Ehsan Sumra, Norman Benedict, Noliel Perez
Cyber-sickness in Virtual Reality: the effect of frames per second and field of view on simulator sickness in VR
Virtual Reality (VR) has great potential in many applications. Researchers are working to create safety training procedures using VR in high-risk industries, as well as surgical simulations and telesurgery. However, there are critical side effects caused by VR applications: disorientation, nausea, and oculomotor function. These virtual sicknesses take away the immersion factor that virtual reality provides. The point of virtual reality is to create an alternate reality for a user made by some developer. It is a huge problem in fields such as medical virtual reality or teaching virtual reality, one cannot stay in these realities for too long due to the sickness they get. These effects can be measured by a simulator sickness questionnaire (SSQ), which is commonly used in VR research. Many factors that may cause simulator sickness have been identified, however, there is currently no graphic specification such as Frames Per Second (FPS) guide to help researchers and developers reduce simulator sickness. We are planning to create a model for researchers and developers to find the lowest SSQ score through the two variables Frames Per Second (FPS) and Field of View (FOV). This will assist developers in finding what may be the best option for FPS and FOV in their module to lower sickness in participants.
Overview of Virtual Reality Simulation Sickness: Causes, Symptoms, and Prevention Strategies
Christa Lehr, Ezequiel Fontan, Xavier Stephens
“Whose Allele is it anyway?” – a proposal for evaluating sequence-based phasing methods
The human genome is a diploid genome, which means there are two copies of each chromosome, except for the X and Y chromosomes. As a results, there are two copies (alleles) of each of the genes laying on these chromosomes. Allele Specific Expression Estimation (ASE) is the problem of estimating the expression level of each gene at the allele level. In other words, it is concerned with finding whether one of the two alleles or both alleles of a gene are actively being transcribed into RNA to generate proteins. If both are expressed (active), then finding out if they are expressed at the same level or at different levels. Transcriptome sequencing is used to address this problem. The analysis starts by comparing the sequencing data to a reference genome. A reference genome is a general genome that is represents all individuals in a species, but is does not completely matches each individual, given that variations between individuals. The reference genome is haploid, including the sequence of one copy of each chromosome. This poses challenges in addressing ASE, where we are interested in identifying difference of expression between the gene alleles coming from the diploid genome. One way to address this issue is through creating a diploid reference from the individual being studied, through first finding where and how this individual genome vary from the reference genome. Then phasing these variations arranges the alleles at different positions into two groups to allow us to create the diploid reference of the individual. In this research project, we will study and compare existing phasing algorithms to select one that has high accuracy and is capable of phasing short insertions and deletions, in addition to single nucleotide variations, which will aid in solving the ASE problem.
Joshua Riznyk, Harry L. Sanders, Siddhi Suresh
Fairness in Facial Classification Algorithms
Our research aims at better understanding how the use of demographically balanced versus unbalanced data sets, in the training and testing of classification algorithms, affects their ability to classify individuals, especially those in minority groups. We plan to utilize untrained classification algorithm(s), training with proportional data (in terms of social categorizations) and disproportional data to determine if algorithmic bias is inherent to the design of the algorithm, or rather the methods and data used in their training. Recent studies illustrate how algorithms are inherently biased even before training, however, we aim to better understand how biased training data can additionally affect the level of bias present within these algorithms. Highlighting this disparity could improve and refine the direction of future studies to assist in the understanding of how these biases infiltrate these algorithms, and in the development of methods to prevent these disparities in future classification models. Open-source, untrained classification model will be trained using both unbiased data sets and biased data sets, with the purpose of comparing the results from each algorithm. By training each algorithm with different data sets and comparing results, a fairness metric will be used to determine how fair these algorithms are, and thus determine the effect of biased and unbiased training sets on the outcome of these algorithms. Additionally, this research plans to draw correlations between the proportionality of the data sets used for training and the quantitative fairness of the algorithm. The fairness metric is defined as the degree to which an algorithm performs equally amongst all subjects, regardless of demographics like race, gender, or age. This metric will use the ratio of the best to worst Positive Predictive values, and from this, it can be determined if there exists a correlation between the fairness levels of the algorithm and the proportionality within the training dataset. This research intends to lead to improved algorithms that, with better accuracy and precision, can lead to better trust and value of this technology and contribute to its fair use in the greater society.
The Effects of Biased Training Data on Intersectional Classification Accuracy in Facial Analysis
James Petkin
James Petkin presented his research in a poster presentation at the 2024 Consortium of Computer Science in Colleges New England (CCSCNE) in Albany, New York on April 12, 2024. James's poster also received Third Best Poster at the Conference. He also presented his work at the 2024 SCSU Undergraduate Research Conference.
Deeper Understanding of Depth Sensing Technology
This research focuses on depth perception and how the current depth-sensing technology can improve image recognition and biometric authentication. Against the backdrop of evolving AI technologies and deepfake practices, this study aims to explore and evaluate the vulnerability in authentication using 3D (with depth) versus 2D images. We employed the StereoBM algorithm from OpenCV to compute disparity maps from pairs of stereo images, mimicking human depth perception. The algorithm introduces a function, extract_2d_features, which utilizes grayscale conversion, histogram analysis, and the Canny edge detector to extract comprehensive 2D features from images. Similarly, extract_3d_features are defined to analyze depth maps through Gaussian filtering, gradient computation, and depth magnitude assessment, forming a multidimensional feature vector. Real-time image processing is demonstrated through continuous frame capture from a camera feed, applying Laplacian filters and Canny edge detection to enhance and visualize edges. Finally, we explored authentication between 2D and 3D images using ORB feature detection and brute force matching in OpenCV, establishing a secure feature comparison and verification methodology. The findings will enhance the understanding of depth sensing technology and inform its secure application in facial recognition and broader implications for the field’s future trajectory.
Samuel Trumbley
Samuel Trumbley presented his research in a poster presentation at the 2024 Consortium of Computer Science in Colleges New England (CCSCNE) in Albany, New York on April 12, 2024. His poster's title is "Predicting Hits with AI". He also presented his work at the 2024 SCSU Undergraduate Research Conference.
Using Artificial Intelligence to Predict the Outcome of Batted Balls and Evaluate the Impact of New MLB Rules
I present an analysis of the MLB’s new rules by training a machine learning model on 2023 batted balls, then predict 2022 batted balls and compare to the actual data from that season. Recently, the MLB has seen certain patterns of change in the sport that have seemingly hurt the viewing experience. The length of games has seen consistent increase, and the sabermetric approach most teams have taken, which values home runs and slugging, has led to less action on the bases and less balls being put into play. The MLB decided to implement several new rules for the 2023 season that would hopefully combat these issues. These rules include a pitch clock, larger bases, and the banning of infield shifts.. In the first season with these rule changes, it seems the MLB has been successful in their attempt to increase offense and make the game more engaging. Statistics like stolen bases, hits, and batting average all showed an increase from the 2022 season to the 2023 season. However, I wanted to dig beyond these statistics to see just how impactful these new rules are, specifically the ban on infield shifts. This idea inspired me to utilize artificial intelligence and machine learning to evaluate the impact by comparing the two seasons. The data from the two seasons were collected from Baseball Savant. Each dataset uses six unique features to predict the outcome of batted balls, and three different classifiers are utilized to project the outcomes. The results from comparing the predicted hits from the 2022 season support the claim that the rules were effective in increasing offense. The predicted hits were greater than the actual number of hits from the 2022 season, confirming that if the rules were implemented in that season, it would have led to more offense.
After participating in SCSU's ERSP Program, Christa went on to a 2023 Summer Research Internship Program at UC San Diego. During her UCSD Internship, she worked with Dr. Christine Alvarado and other researchers, and their work resulted in an ACM SIGCSE publication titled Understanding California's Computer Science Transfer Pathways.
Upon Christa's return to Southern's campus, she was awarded a SCSU Innovation Hub research scholarship working under Dr. Sahar Al Seesi. Christa shared this work in a poster presentation at the 2024 SCSU Undergraduate Research Conference.
This project is made possible by a sub-award from University of California, San Diego and through the support of the National Science Foundation under Grant No. DUE-1821521 and a Google Initiative.