LI, Mingyang

About Me

Background

Mingyang Li received two Bachelor's degrees studying nanomaterials. Just when he thought he'd create his own self-healing superhero armor, he found himself in love with data crunching more than with magnetron sputtering¹. Following his heart, soon after joining the Scientific Computering (SCMP) program at UPenn, he submatriculated into DATS.

Work

I am a Research Associate at Wharton Research Data Services (WRDS), where I build pipelines to parse SEC filings (Form 10-K, etc.). I also conduct social science research at the World Well-Being Project (WWBP), mainly focusing on comparing Chinese and American cultures using microblog data. I am joining Google as a software engineer in late September, 2019.

Publications

• Guntuku, S. C., Li, M., Tay, L., & Ungar, L. H. (2019). Studying Cultural Differences in Emoji Usage across the East and the West. Proceedings of the International AAAI Conference on Web and Social Media, 13(01), 226–235. Slides.
• Li, M., Guntuku, S. C., Jakhetiya, V., & Ungar, L. (2019). Exploring (Dis-)Similarities in Emoji-Emotion Association on Twitter and Weibo. Slides.
• Li, T., Li, M., & Huang, J. (2016). Quantum Fisher information of triphoton states. Chinese Optics Letters, 14(3), 032701. DOI: 10.3788/COL201614.032701. PDF.

Course Notes

Shared here are my review notes and summaries for some courses I have taken. Hope you will find one helpful.

CIS545 Big Data Analytics - Spring 2018

This was a fun course to attend.

• Term Project Report: Titled "A Quantitative Study On How Emojis' Meanings Shift From West To East," this report innovatively used radar chart to compare how each emoji projects (as in linear algrebra) to five of human's basic emotions (see Ekman’s List of Basic Emotions, 1972) across English-speaking American users on Twitter and Mandarin-speaking Chinese users on Weibo.
• Course Notes After Midterm: All course notes taken in-class and after-class (while watching again the recordings) for the second half of the course. Source.
• How to set up "All-Spark Notebook" on AWS EC2 Cloud Service: A tutorial on how to set up the docker image "All-Spark Notebook" -- a painless way to get started with this course -- on AWS EC2 Cloud Service.
• How To Set Up Docker At SEAS: A tutorial on how to set up docker containers inside a virtual machine that is run (possibly remotely) on a SEAS server.
• K-Means In SQL: SQL statements of two implementations of the K-Means Clustering Algorithm.

STAT512 Mathematical Statistics - Spring 2018

Taught by Professor Ewens, this course was a charm.

• Know For Midterm 2018: Formulae and takeaways. Source.
• Map of Concepts and Map of Topics: Scanned document of two hand-drawn maps.
• All Course Notes: Complete course notes does this course since Day 1. Hand-written; digital. ~105 MB.
• Some Formulae And Theories: Probably sufficient for the final exam.
• Lambda Ratio In ANOVA: A hand-written page detailing how we came to the F distribution deriving from the lambda ratio for the ANOVA example. This is perhaps the most general case we could consider for normally distributed random variables.
• Lambda Ratio In Regression: A hand-written page detailing how we came to the T distribution deriving from the lambda ratio for the linear regression example.

CIS550 Database - Fall 2017

• Cheatsheet Academic Integrity Sheet For Final Exam: If a "cheatsheet" is officially allowed during an exam, why it's called a CHEATsheet?

CIS502 Algorithm Analysis - Fall 2017

I got a C for this course. Don't believe anything I wrote in my notes.

• All Algorithms: Every algorithm encountered during this course. Source.
• How To Properly Answer Questions: Lacking formal former education in computer sicence, I found myself in need of explicitly writing down how I should phrase myself for answering many questions. Source.
• NP-Complete Reductions: Diagram of how many problems can be reduced to other problems. Source.

CIS519 Intro To Machine Learning - Fall 2017

Notes here are in Markdown format. I recommend Typora as your Markdown editor/viewer.

• How can SVM maximize the margin in decision boundary by minimizing the \sum_{j=1}^d\theta_j?: A stub. Source.
• Midterm Review Notes: Carefully crafted were the table comparing many algorithms to construct machine learning models and a description of problem setting of regression models. Source.
• How To Run Spyder On BigLab: For Penn Engineering students with BigLab access only. A tutorial of how to run Spyder on BigLab servers but display it like a native window on your computer. Source.
• How To Make Your Life Easier With CIS519: For Penn Engineering students only. A tutorial for non-tech-savvy students to get started with this course without ruining your own computer. Source.
• Final Review Notes: you may find interesting the following topics: (Source)
  1. General procedure of training and evaluating a model.
  2. For each of many machine learning models, may it get trapped at local minima?
  3. For each of many machine learning models, how biases and variances change with increasing parameter.

MNS321 Electrical and Optical Properties of Materials - Spring 2016

Taught by Professor Gunter Scholz at the University of Waterloo -- yes, I was a nanomaterial student before diving into data science.

• Summary: Many concepts summarized into a single document.
• Formulae And Derivations: As title suggests.

Miscellaneous Notes

Listed here are notes not related to any course and also not suitable as a blog post.

• Assets (Finance): A mind map of some important concepts involved with assets as a financial term. Mostly copied from Investopedia and Wikipedia. Source.
• Mindmap of linguistics, as a field of study: My brief understanding of the relationships between subfields under the realm of linguistics. Source.
• How To Install TensorFlow with GPU Support on Debian: Your mileage may vary. Source.
• Classification Of Magnetic Materials: A table that compares some types of magnetic materials. Source.
• How to Take Notes Effectively in 2017 (for University Students): A mind map that serves as the draft of a blog post I planned to finish by 2018. Source.
• Revenue Is A Poor Indicator For Measuring A Tax-Haven-Based Subsidiaries’ Contribution To Tax: A practicum report.