Quantifying Noise in Dimensionality Reduction Methods
Computing the noise propagation in t-stochastic neighbor embeddings (t-SNE) and uniform manifold approximation projection (UMAP).
Online Adaptive Controller Selection for the Stable Control of Noisy Dynamical Systems
Proving that model predictive control and regularized online balanced descent with noisy predictions can control dynamical systems.
Enlargening the Pseudorandomness of Expander Random Walks
Finding classes of test-functions that are fooled by random walks on expander graphs.
Identifying Chemicals in Metabolomic Samples through Dimensionality Reduction
Identifying Chemicals using Mass-Spec data, aided by t-stochastic neighbor embedding (t-SNE) and Nonnegative Matrix Factorization (NMF).
Correlating COVID-19 outbreaks to COVID-19 Vaccine Efficacy
Using financial machine-learning models to quantify COVID-19 cases as a function of vaccination rates to predict outbreaks.
LIGO: Acoustic Invisibility Cloak
Using particle swarm optimization and finite element analysis to design/simulate acoustic band-limiting filters for the gravitational wave detector's Cryo-Q upgrade.
Ph11 Hurdle 1: Modeling Airplane Delays at SFO
Creating a mathematical model for the delays of airplanes at the San Francisco Airport during the closure of runway 85A.
Ph11 Hurdle 2: Modeling Rinderpest and Wildebeest Populations
Modeled Wildebeest, Grass, and Rinderpest virus populations in the presence of forest fires using analytical factors of the Leslie Matrix.
Hidden Markov Model to Generate Shakespearean Sonnets
Created an unsupervised Hidden Markov model using the Viterbi algorithm to generate links between word-phrases, and trained the HMM to generate novel Shakespearean sonnets.
Piano Synthesizer: Java
Implementing a music synthesizer on a piano, with a Karplus-Strong algorithm to modulate frequency resonance and amplitude decay.
Optimized Dynamic Memory Allocator
Created an optimized dynamic memory allocator using an explicit free list with coalescing and block-splitting techniques to enhance the utility and speed of the allocator.
Tim and TCompiler
Used node-extraction and parsing to create an assembly language generator (TCompiler) for a primitive programming language (Tim) that can perform complex computations.
Is Ditch Day Tomorrow?
A webpage that predicts whether tomorrow will be Caltech's ditch day. The probabilities used here are based on the past 20 ditch days.