Kayla Irish
Kayla has been a data scientist in the Wheeler Lab since she graduated in December 2021. She was previously an undergraduate researcher for the lab, starting in December 2020. She holds degrees from the University of Montana in History and Mathematics, with a concentration in statistics. In the fall, she will begin her PhD in Statistics at the University of Washington to study machine learning and ethical AI. She recently discussed her interests in this Senior Spotlight created by UM.
In the lab, Kayla is currently finishing a joint research project with Doug Emlen’s evolutionary biology lab. DISCO (DISCO implements Sound Classification Obediently) combines 1-D CNNs, U-Net, ensembling, and Hidden Markov Models to automate the classification of sound files. With the Emlen lab, this work aids in automating the labeling of rhinoceros beetle courtships, but the tool generalizes to any recording. This is a sample beetle courtship recording spectrogram with automated labels below it:
The green areas in the spectrogram are classic “B” chirps. The first bar below it, labeled “ensemble prediction,” shows the machine learning ensemble predictions for each timepoint in the spectrogram. Here, purple indicates a “B” chirp classification, red indicates an “A” chirp (absent from this image) classification, and gray indicates a background noise classification. The next bar shows these predictions after they are processed by a Hidden Markov Model combined with other heuristics. This work will appear in a preprint soon.
Kayla recently won the University of Montana Humanities Institute’s Richard Drake Writing Award for her history research essay, “‘A New Torch in the Cold War’: The Tibetan Uprising and American Religion and Rights Politics, 1959-1960,” advised by Kyle Volk. In statistics, she is a co-author of Newcomer et al., “Identification of geospatial clusters of undervaccination patterns in early childhood using immunization information system data,” which has been submitted for review in Spatial and Spatio-temporal Epidemiology. She also recently finished a consulting project supervised by Jon Graham assessing whether interventions in Billings Clinic caused changes in heart failure patient readmission rates, length of stay, and in-hospital mortality. This paper will be submitted for review by the end of the year.
Kayla is originally from Lewistown, Montana. In her free time, she enjoys cooking, embroidery, cycling, and hanging out with cats.
Pictured above is Kayla with her co-workers.