Math projects
-
Markov chain Monte Carlo
I am working with professor Michael Satz on Markov chain Monte Carlo (MCMC), specifically in Bayesian models. We look at different Monte Carlo sampling methods, such as Hasting-Metropolis algorithm, to see how well they converge to the target distribution and test out ways to increase MCMC's computational efficiency.
[repo] -
Tension Spline Interpolation
In my numerical analysis project, my group developed solutions for tension spline interpolants. Given a number of data points, interpolation estimates the value of an underlying generating function for different values.
I developed and implemented an algorithm to derive a system of equations as a tridiagonal matrix then solved for the coefficients, and wrote the Numerical Computations chapter of the report. This project gave me a deeper understanding of estimating complexity and error for interpolation, and choice of technique. I also investigated what approximation or polynomial expansion techniques that base libraries such as GNU C or LAPACK use underneath.
Math 212 Numerical Analysis: Tension Spline Interpolation [pdf]
Social Computing
-
Predicting Emotion and Attention from Smartphone Behavioral Data
During summer 2021, I worked with Professor Jeff Huang at Brown University with his HCI lab. I worked on Automotion, software-hardware platform for running remote usability testing for mobile devices.
I modified, retrained models to predict users' affect from finger pressure and 6dof motion replay. I reviewed literature on affect prediction using mobile computing and rewrote the paper for resubmission.
Predicting Emotion and Attention from Smartphone Behavioral Data
Dinithi Silva-Sassaman, Cat Mai, Jing Qian, Jeff Huang
Clark Fall Fest '21 (poster) [pdf]
-
Investigating Drug Addiction Discourse on YouTube
By analyzing 166 YouTube videos, we found that YouTubers leverage videos to disclose personal addiction experiences, provide professional recommendations, and express addiction-related opinions and lifestyles. I watched and tagged videos using grounded theory, and discussed with the group to extract common themes.
Investigating Drug Addiction Discourse on YouTube
Katherine McKim, Cat Mai, Danielle Hess, Shuo Niu
CSCW '21 (poster) [pdf]
-
#TeamTrees: Investigating How YouTubers Participate in a Social Media Campaign
#TeamTrees is a year-long 2019 initiative that managed to raise $20 million before 2020 to plant 20 million trees. Using YouTube's metadata, we investigated how different category of content creators participated in #TeamTrees.
I crawled the data set and created the video-tagging interface. I researched #TeamTrees origination and impact, as well as literature on civic engagement and grassroot movement building. I watched and tagged more than 500 videos and discussed them with other authors. I ran the analysis and wrote parts of the paper.
#TeamTrees: Investigating How YouTubers Participate in a Social Media Campaign
Shuo Niu, Cat Mai, Katherine G. McKim, D. Scott McCrickard
CSCW '21 [pdf]
-
#StayHome #WithMe: How Do YouTubers Help With COVID-19 Loneliness?
We used Weiss's loneliness theory to examine how YouTubers provided friend-like, mentor-like, and family-like provisions through videos in different styles during COVID-19.
I crawled the dataset for the project and automated the data pipeline. The whole lab then participated in ground theory analysis to come up with the provision and video styles specific to this campaign.
#StayHome #WithMe: How Do YouTubers Help With COVID-19 Loneliness?
Shuo Niu, Ava Bartolome*, Cat Mai*, Nguyen B. Ha*
CHI '21 [pdf]
-
Emotional Analysis Via Content Creation on YouTube During COVID-19
This is my final project for my data science class with Nguyen Ha. We were interested in the emotional response and coping mechanism of content creators on YouTube during COVID-19.
We applied the fight-or-flight framework to analyze how content creators could contribute to mental health for audience. Our approach was to analyze the description and hashtags of the video and use Naive Bayes classifier to create a predictive tagging model.
[final report]
Other
-
YouTube Scraper and Decoder Python Package
Dr. Shuo Niu and I wrote the beginner-friendly YouTube crawler pipeline and a .json decoder specifically for retrieved file from YouTube API request. I packaged it and am currently the lead maintainer. My lab has been using this package for several projects.
[stable release (limited docs)] -
Geovisualization of Garden of Bomarzo
During summer 2019, Dr. John Garton and I were building prototypes locally and developing proof of concept of a virtual garden visit in efforts to preserve the historical site.
I used Agisoft Metashape to render 3D models based on photos from Garden of Bomarzo. My early attempts at processing the model of the statues helped Dr. Garton’s colleague to secure a national Futures grant and helped Dr. Garton to present the project to grant boards.
[demo] [news]