Page menu

Hidden Markov models

Gabrielle Warren uses R to explore HMMs (2017 senior projects, part 5)

Gabrielle Warren, a senior, majors in mathematics and chemistry, presented a poster about her work during the Mathematical Association of America sectional meeting at Washington State University.

This seven-part series highlights academic projects completed by Walla Walla University students during their senior year.

For her senior project, Gabrielle Warren, who will graduate in December 2017 with majors in mathematics and chemistry, researched and implemented a low-dimensional hidden Markov model (HMM). Last school year, after learning about the programming language R, Warren started using it to duplicate HMM simulations. The connections between mathematics and chemistry inspired her to use an HMM in R to explore the mathematics behind computational drug discovery—a field of chemistry that uses various molecular modeling techniques to predict potential drug candidates.

HMMs are used when underlying causes are not directly related to observable outcomes. They are used in bioinformatics, speech recognition, and finance in particular. In order to better understand HMMs, Warren generated a dataset for testing a low-dimensional HMM. The dataset included a set of underlying states and a set of simulated observations. Comparison of the predicted states to the original set of states provided a standard for the evaluation of the Viterbi algorithm.

Even as a top mathematics student, Warren found the project challenging because of the dense nature of the literature written about HMMs and the need to continuously debug the code she was using to test the model. After all the hurdles and pitfalls, it was very rewarding to run the code and see the results. While working on the project, Warren also found that HMMs are used in biology more often than in chemistry.

Although her project started simply as an academic exercise, her work has potential practical uses as well. For example, using the programming language R to generate the model means that the code is easily manipulated and students in bioinformatics or finance classes could modify the code to model their own data. Many biologists are familiar with the language R and use HMMs, so her project could help both areas become more accessible to students.

In May Warren presented a poster about her work during the Mathematical Association of America sectional meeting at Washington State University that was organized by WWU alumnus Kevin Vixie. She has plans to attend graduate school in chemistry.

Posted Aug. 23, 2017

Last update on November 23, 2015