Derrick (Eui Gyu) Kim

NYU Dean's Undergraduate Research Fund recepient(September 2018 ~ May 2020)

-Implemented K-Nearest Neighbor(KNN) algorithms from scratch on breast cancer cell data set and least squares for classification on a manually created data set.

-Implemented Naive-Bayes Algorithm on complex data such as blood cell images and simpler data such as images of handwritten digits(0-9).

-Studied the theoretical mathematics behind Support Vector Machines(SVM) and Naive Bayes Algorithm.

-Experimented with Interior Point Method(IPM) and the challenges involved if the number of input points is scaled up. On a training data set that has binary labels, formulated a Support Vector Machine Quadratic Program, using Karush-Kuhn-Tucker(KKT) conditions to convert the program into a linear system.

-Used a log barrier to ensure that the iterations of the IPM problem stay within the feasible region.

-With the newly evolved constraints, attempted to find the optimal hyperplane that classifies the modified problem.

-Submitted my thesis to Professor Sondjaja for review.

Solving SVM using IPM PDF | CODE
Derrick Eui Gyu Kim

Graduate Research Assistant at USC Information Science Institute

Timeline: (January 2022~ Present)

Analysis of Deepfake Detection Algorithms

-Analyzed Deepfake Detection Algorithms by lowering the resolution of images in Deepfake Datasets to see if there are any common clues or features to detect Deepfakes. If there are such features, then there would be a more robust approach to detect such fake images.

-Explored large datasets by applying models such as ResNet, MesoNet, and Xception-Net.

-Experimented with using adversarial training to improve model robustness and clean data accuracy.

-Retrained the models using darkened images perturbed with Gaussian noise.

Study of Machine Vision Robustness

- Analyzed Human Visual Contrast Sensitivity by creating a new framework based on model architecture and model size to measure the perception ability of machine vision beyond total accuracy

- Attempted to address shortcomings in machine vision in terms of contrast sensitivity.

- Curated datasets CIFAR10–F and CIFAR100-F based on foreground colors, and CIFAR10-B and CIFAR100-B based on background colors, and these datasets acted as a proxy to understand color vision.

- Mirrored the adaptations of human vision to ambient light by using corrupted images from the CIFAR-C dataset.

- Showed how color differences affect model perception in CNNs, and we did further analysis of this effect on various model architectures, model sizes, corruptions, and augmentations and demonstrated that there is a reliance on contrast in CNNs.

- Submitted a paper to CVPR and under review - third author

On Human Visual Contrast Sensitivity and Machine Vision Robustness: A Comparative Study PDF
Ming-Chang Chiu, Yingfei Wang, Derrick Eui Gyu Kim, Pin-Yu Chen, Xuezhe Ma

Old SAT:

2250/2400

TOEFL:

116, but fluent in English as I was raised in New Jersey and spent my undergraduate in New York.

GRE:

Verbal 161, Quant 165, Writing 5.0

Proficient in:

Python(Expert), Java(Expert), C++(Intermediate)

Tools that I am familiar with:

Pytorch, Git, Matplotlib, Seaborn, Unity, OpenGL

Languages:

Korean(Native), English(Native)

Projects for Relevant Coursework

1. Artificial Intelligence(NYU)

-Implemented Davis Putnam Algorithm to generate a Hamiltonian Path

-Created a program that can solve the Post Correspondence Problem

2. Foundations of Artificial Intelligence(USC)

-Implemented Breath First Search, Uniform Cost Search, and A* Search algorithm

-Devised an agent that can play GO on a miniature 5x5 board using the Alpha-Beta Pruning algorithm

-Implemented a Neural Network from scratch to classify the handwritten digits in the MNIST dataset

3. Analysis of Algorithms(USC)

-Wrote a working solution of the Sequence Alignment Problem in python

4. Advanced Mobile Game Design(USC)

-With a group of coworkers, I coded a working game in Unity called "NFT Tinder", which is a satire on the modern obsession with NFT

NYU Presidential Honors Scholar(September 2016 - May 2020)

-Participated in a Scholar Seminar in the first and second years that provided guidance and faculty mentoring focused on developing and building the skills needed for research at the honors level.

-Participated in Advanced Honors Seminars in my second and third years.

-Applied for a grant from the Dean's Undergraduate Research Fund(DURF) and received $1000.

-Achieved a 3.65+ GPA.

-Completed the Honors track in CS major.

-Wrote a thesis prior to graduation.

-Participated in the Scholars Lecture Series

-Visited one of NYU Global Network in Madrid with fellow scholars.

NYU College of Arts and Science Future Scholarship Recipient(2016-2020)

- Received average of $10000 per academic year due to a mixture of financial need and good academic standing

NYU Dean's Undergraduate Research Fund Recipient(2020)

-Received $1000 for my Senior Thesis

NYU Housing Office Web Development Intern(2016-2020)

-Worked 20 hours a week for 4 years

-Assisted clients via email and telephone for housing reservations, transfers, and deadline reminders

South Korean Military(2014 - 2016)

-Was a squad leader for 11 months

-Was a telecommuncations officer for 17 months, and was trained to set up 10 feet antennas efficiently(under 30 minutes)