Signal Detection with Neural Networks and Optimal Detectors (advisors: T. Patrick Xiao and Christopher Bennett)
As a graduate research and development intern at Sandia National Laboratories in summer of 2023, I designed simulations to compare the performance of various methods for signal detection. Specifically, I compared theoretically derived optimal signal detectors against neural network based approaches where I demonstrated that neural networks approach the performance of optimal detectors given sufficient training data.
Nonsense Correlations (advisor: Daniela Witten)
As a statistical consultant for the UW Center of Excellence in Neurobiology of Addiction, Pain, and Emotion, I helped develop methods to conduct valid inference in neuroscience experimental trials. Specifically, many of our problems suffered from “nonsense correlations” where due to autocorrelation structure in time-dependent data, p-values for assessing correlations in data are deflated: leading to so-called “nonsense correlations.” Our research investigates methods to avoid nonsense correlations using by generating an approximate null distribution of the test statistics.
Ensemble Kernel Density Estimation (advisor: Kevin Moon)
Kernel density estimation is a nonparametric technique for estimating a probability density function from a finite sample of data. As an undergraduate student at Utah State University, I helped derive an ensemble-based approach to kernel density estimation which in simulations empirically improves over using single kernel density estimators.
Autocorrelative Regression Trees (autocart)
Regression trees are a popular machine learning method for regression and classification. I have researched modifications to the regression tree algorithm such that it improves results on spatial datasets featuring coordinate information.
Ethan Ancell 