Highlighted Project

The pandemic worsened eating disorders among adolescents, particularly girls. Eating disorders, a psychiatric illness where an adolescent tries to control their weight with severe food restriction (anorexia) or purging (bulimia), can be fatal. How much does web content and algorithms that power web search contribute to eating disorders? Imagine a girl who is unhappy about her weight and looks for dieting tips online. It does not take long for search algorithms to lead her to extreme content that promotes anorexia and extreme weight loss. The goal of this project is to audit the web for potentially harmful content and communities that promote eating disorders.

There is anecdotal evidence that broadband service offers vary in quality and price along income and racial lines. This project seeks to validate these claims by collecting data about service speeds and prices for all known serviceable locations in Los Angeles County, and merging with sociodemographic variables from the Census Bureau and other sources. Other researchers have already developed prototype code for scraping data from ISP’s websites, but the study only covered a small percentage of addresses in LA County. The analysis will probe for evidence of a “poverty penalty’ whereby residents of poorer areas in Los Angeles are offered higher cost/lower quality broadband services.

Analytics to determine what tertiary and quaternary care patients would have the best outcomes at Keck Medical Center. Analyze data from incoming transfer patients to look at data quality, evaluate clinical records and predict outcomes.

The human corpus callosum is the largest pathway connecting the left and right hemispheres of the brain. The shape of the corpus callosum (CC) changes throughout the course of human development, and it can also be altered with respect to disease onset. We can explore the variation in CC shape along the middle of the brain, but we need to extract it reliably first. The lab currently has two methods for extracting the CC, one using only image processing techniques, and another using deep learning (UNet) but these methods do not always extract the CC accurately. The accuracy results often depend on the MRI scanner that was used, or the abnormalities present in the scan. Can we improve the performance of our deep learning model with additional training data? Can we change some processing steps to improve the model? Once we do have an accurate segmentation, then what shape metrics of the CC as a whole, or in parts, are most telling of the underlying biology, such as age and risk for disease?

This new project attempts to understand the interaction between anti-vaxxers (anti-vaccination groups) and alt-right groups on platforms such as Facebook. The goal of this project is to understand how do these two types of fringe groups interact over the years, and how do their interactions and discourse evolve during the COVID-19 pandemic. It would be interesting to explore the longitudinal patterns of network/discourse co-evolution and how such patterns may change in times of dramatic events.In terms of data, I have access to Facebook’s historical data archive and I have collected a dataset that contains a list of anti-vaxxer (n=158) and alt-right groups’ (n=183) Facebook posts over 10 years (2010-2021). The dataset can be further expanded with additional help.

College football recruiting is big business. This project aims to build and analyze a comprehensive college football recruiting dataset, to help determine if there are biases in who and how college football coaches recruit players. This data set will combine college football recruiting data from the web with census and other socioeconomic data, to search for patterns in where and how college football coaches recruit players.

Common sense knowledge bases are used widely in research, spanning many areas in artificial intelligence, including natural language understanding, computer vision, and planning. However, these resources may contain human biases, which will ultimately be embedded in the resulting AI solution and potentially have negative societal impacts. The extent to which these biases exist is unclear. In this project, you will define several well-motivated biases (location, gender, ethnicity) and measure the extent to which they are represented in ConceptNet.

Developing models to analyze images is a demanding task that requires significant time, resources, and effort. Recently, companies such as Amazon and Google are providing services to make the modeling process easier so even users with little machine learning expertise can enjoy deep learning technologies. Based on our prior work in object detection and classification for smart city applications, we would like to compare and evaluate the process and performance of commercial services using our training datasets. This project will be a good practice to understand the image machine learning modeling process and the advantages/limitations of commercial services for customized learning.