Computer Scientist, Mathematician, and Software Engineer interested in Machine Learning and pushing the boundaries of understanding in Computer Science. I have been programming, tinkering, and constantly curious for over a decade and I consider it to truly be my lifelong passion. I am mathematically obsessed and still exploring the realm of academia & research, but I find the intersection of Computer Science and Mathematics to be fascinating. My current research interests lie in solidifying the mathematical underpinnings of deep learning. I work as a Data Scientist & Software Engineer at Hinalea Imaging working with novel machine learning-based sensing on hyperspectral imagery.
I lead the effort to develop state-of-the-art hyperspectral machine learning algorithms at Hinalea Imaging, a TruTag Technologies subsidiary focused on the development of hyperspectral imaging and sensing.
Contributions:Built a bi-directional sync engine for "configuration as code" in .NET C# for the Microsoft 365 platform -- running completely on serverless architecture.
Developed automatic version control for detection and resolution of state changes detected in users’ Microsoft infrastructure configuration.
Was also responsible for building out a user interface on the web in Vue, Typescript, and SASS.
Studied embedding and visualization of high-dimensional and non-Euclidean data using unsupervised and supervised Deep Learning networks under the supervision of Professor Shuyang Ling. Careful consideration was made in understanding the mathematics that underpin these techniques.
Authored the Senior Capstone Thesis “Exploring the Limitations of t-SNE” under the supervision of Professor Shuyang Ling. This work aimed to find of the limits of popular dimensional reduction and visualization techniques, such as the t-SNE dimensionality reduction algorithm. The manuscript specifically explored degenerate embeddings of high-dimensional data, applications to latent geometric processes in image sequences, and unscrambling randomized Radon Transform tomography data.
Researched emerging post-quantum, lattice-based cryptography schemes under the guidance of Professor Siyao Guo of NYU Shanghai, as well as delving into the inner workings of quantum computing theory. The goal of the research was to investigate some of the many open questions in post-quantum cryptography and work toward answering them.
Gained experience in writing proofs for proving the security of cryptographic protocols, which are critical in ensuring the security of existing and emerging schemes. Left to resume pursuing interests in Machine Learning.
Institution | Degree | Dates Attended |
---|---|---|
NYU Shanghai | B.S. Computer Science | Sept 2016 - Dec 2020 |
NYU Shanghai | B.S. Mathematics | Sept 2016 - Dec 2020 |
Until GitHub provides a better API for showcasing repositories, you can find my recent and pinned repositories on my GitHub.
Unfortunately, all I have to show here right now are papers I wrote in my naive undergraduate research phase.
Morlock, Frederick. "Exploring the Limitations of t-SNE" NYU Shanghai Senior Capstone Thesis (2020). Available Here .
I wrote my undergraduate Senior Thesis paper "Exploring the Limitations of t-SNE" under the supervision of Professor Shuyang Ling of NYU Shanghai. The paper experimentally analyzed 3 problems using the t-SNE algorithm to test both its limitations and possible novel applications. The paper explored the effects of the "crowding problem" when embedding high-dimensional data in 2D, the preservation of latent geometric structure in embeddings, and the application of unscrambling the "Scrambled Radon Transform" in tomography data. As much as I regret it, while the paper was theoretically motivated, there was not a rigious analysis of the phenomenon, but rather just empirical studies.
Morlock, Frederick. "Graph Embedding and Visualization Using t-SNE" Deans’ Undergraduate Research Fund -- NYU Shanghai (2020). Available Here.
I received funding through the Deans' Undergraduate Research Fund at
NYU Shanghai, where I conducted research under the supervision of
Shuyang Ling. In the manuscript "Graph Embedding and Visualization
Using t-SNE", I furthered some of the research I did as part of my
Thesis and explored the application of t-SNE to graph embedding and
visualization. In the paper, I studied applying t-SNE for use in graph
embeddings. I primarily looked at 3 datasets: Facebook's "Egonet"
dataset, the "JAGMESH" mesh network dataset, and a simple "Stochastic
Block Model" network. While t-SNE possesses the ability to be able to
perform graph embeddings, one of its limitations is the computational
complexity required to compute it. There were 3 approaches that were
taken to overcome this limitation: graph sparsification via "effective
resistance" approximations, geodesic distance approximations via the
Nystrom Approximation, and graph coarsening via Kron Reduction. While
Kron Reduction was viable in some sense for reducing the computational
complexity of the problem, given that we wanted to preserve all of the
nodes in the original data in our embedding, I did not include results
using the Kron Reduction as graph coarsening (removing/combining graph
nodes) as it was not in the spirit of the paper's objective.
It may be noted that this paper was written and submitted in 2020,
despite the linked paper header reading 2021. The header there
contains the date when I compiled that version of the paper since I
had my LaTeX heading set to read the current date...
Morlock, Frederick, and Dingsu Wang. "MAD-VAE: Manifold Awareness Defense Variational Autoencoder." arXiv preprint arXiv:2011.01755 (2020). Available Here.
Personally, I'm not sure that anyone should read this paper. It was from a period of my life when I was naive about research, and it was published long after it was written. Not only was I a naive young aspiring researcher, but the study of adversarial examples was young at the time as well. While I haven't studied this area for a while, if you are curious about adversarial examples, then you might want to read the paper "Adversarial Examples Are Not Bugs, They Are Features" by Ilyas et. al.
You can reach me on any of the following social media platforms. Additionally, you can find my email in my resume.