Machine learning/artificial intelligence, time series analysis & prediction, biological physics, model development and validation
Techniques Brad Parry Uses:
Programming in Python & Tensorflow
Machine learning & AI algorithms
Classical statistical inference
Model validation and robustness testing
Brad Parry's Problem Solving Skills:
Evaluating model quality and robustness
Developing novel machine learning algorithms for new problems
Applying deep and reinforcement learning
Brad Parry's Problem Solving Experience:
I developed novel machine learning algorithms to measure physical properties of the bacterial cytoplasm. This project began with development of an experimental strategy to gather large quantities of single particle dynamics from living bacterial cells. To decipher the experimental data, I developed novel machine learning algorithms to identify structure in the data. As a result, I uncovered new types of non-linear behavior not previously known in the bacterial cytoplasm.
I Devised an unsupervised machine learning approach to help Mongolian anthropologists identify areas of human habitation in an ancient site.
I wrote programs to analyze metabolic pathways in cold-adapted species. I subsequently applied these findings to warm adapted bacterial species and successfully increased their low temperature growth rates and cold tolerance.
I performed frequency decomposition of macroeconomic time series and assembled the decompositions with boosting and ensemble methods to build a model that predicts the yield change of the 10-year US Treasury over 3 month time periods.