I am interested in applying machine learning methods in the context of scientific problems and the unique challenge that poses: how do we learn and assess performance when uncovering the ‘ground truth’ is the objective? I take an engineering approach, starting from the specific problems of interest in the data, then modeling them and selecting appropriate performance measures to understand the performance of the model in the context of the problem.

My technical background is at the intersection of signal processing and machine learning, and I currently collaborate with psychologists, analyzing data from psychological experiments.


Project Summaries:

More details and related products are on the individual project pages.

A Sparse Combined Regression-Classification Formulation for Learning a Physiological Alternative to Clinical Post-Traumatic Stress Disorder Scores

An applied machine learning AAAI-15 paper on a method for learning a scoring function for PTSD diagnosis from peripheral physiology measurements that maintains important properties without being subject to the same weaknesses of the ‘gold standard’ score produced by a clinical interview. We present the problem as a slightly modified learning task from a typical setting, define desired properties of the method and propose a sparse combined classification-regression loss function for learning.

Machine Learning Analysis of Peripheral Physiology for Emotion Detection

My master’s thesis work used machine learning techniques to analyze physiological data for detecting discrete emotions. This work applied temporal models to data collecting during experiments where subjects were shown emotionally evocative images and sounds.

Comments are closed, but trackbacks and pingbacks are open.