AI & ML
Microstructural Classification of Unirradiated LiAlO2 Pellets by Deep Learning Methods
(With co-authors Karl Pazdernik and Nicole L. LaHaye). Recognition and characterization of microstructural features is essential to the understanding and prediction of material performance under various operational conditions, including irradiation. In this work, we tested a collection of Deep Convolutional Neural Network (DCNN) architectures that have been optimized for image segmentation and selected the best performer to obtain pixel-level classification of the main microstructural features in unirradiated LiAlO2 pellets, including grains, grain boundaries, voids, precipitates, and zirconia impurities.
Handling Unkown Transition Dynamics in REstless Multi-Armed Bandits
This is my review paper for my written prelim exam and a brief simulation paper. In the review, I compare and discuss Q-learning RMAB algorithms. In the simulation study, I compare basic Q-learning RMAB algorithms varying levels of heterogeneity among the arms and varying levels of budget constraint. Click the link below for a combined PDF.
THompson Sampling in Multi-armed bandits for treatment assignment under spillage effects
This is my class project with Kyle Duke for our PhD Bayesian Statistics course. We compare online MAB algorithms in a clinical treatment assignment setting on synthetic and real datasets to find out which algorithm is most robust to treatment spillage effects among treatment arms. Final report available.
Considerations for incorporating ai with human feedback in airport risk assessment models
Jonathan Mills, Rafael Ferreira, and I wrote this report while interning at Pacific Northwest National Laboratory. We examine the problem of creating an automated parameter tuning method for optimizing an airport risk assessment model (ARAM) that minimizes a risk objective by allocating resources across an airport and adapts based on streaming event data.
Open ABm GYM: Fast & Replicable prototyping for Agent-based models
After spending over a year collaborating with social scientists and computational specialists, I wrote a Python package that wraps MESA to make agent-based modeling accessible for entry-level Python users while providing speed and ease for common needs in agent-based modeling for computational experts.