Patrick Herbert Email and Phone Number

Developing machine learning algorithms that allow for sensitive and specific identification of patients at risk of undergoing rapidly progressing glaucoma.

- Implemented mixed integer linear programming (MILP) to evaluate deep learning model vulnerabilities and developed a pruning algorithm to identify critical neurons.- Applied adversarial training techniques and MILP formulations to compare original and pruned networks, revealing unexpected improvements in both accuracy and robustness. Utilized confusion matrix analysis and created visualizations using matplotlib to present findings.

- Imported and assessed fMRI scans of human test subjects for a series of projects using Python- Trained a model on limited data using cross-validation in order to predict neural representations when given a target word- Applied Monte Carlo permutation tests to evaluate statistical significance between the model's predictions and the actual data- Produced data visualizations to aid in the presentation of our results