05/04/2026
Congratulations to Lindy Hyde and Hannah Baird, who were both selected to present their research posters at the 2026 Pacific Inland Mathematics Undergraduate Conference (PiMUC). PiMUC is a conference featuring presentations from Pacific Northwest undergraduates on math, statistics, and data analytics research. Below, are the abstracts for their presentations:
Lindy Hyde (mentored by Dr. Maryam Bagherian):
'Improving Classification with Enhanced Convolutional Neural Networks via Knowledge Distillation'
High-dimensional and noisy datasets pose challenges for reliable classification, particularly in applications such as neuroimaging and facial recognition. In this work, we investigate an integrated machine learning pipeline that incorporates preprocessing steps (including denoising and data augmentation) and combines convolutional neural networks (CNNs) with encoder–decoder architectures and support vector machines (SVMs). To address class imbalance and data heterogeneity, we employ a multi-teacher knowledge distillation (KD) strategy, where multiple teacher models, each trained on different subsets or aspects of the data, provide guidance to a student model. This setup aims to transfer complementary information and improve performance on underrepresented classes. Experiments on neuroimaging and facial recognition datasets suggest that this combined approach enhances classification consistency and achieves competitive accuracy compared to baseline methods.
Hannah Baird (menored by Dr. Derek Eckman):
'Winnowing the Pool: Screening and Selecting Participants for Qualitative Interviews'
Recruiting qualified participants for a qualitative research study requires more than just a generalized community announcement asking for volunteers. Determining a group of participants with diverse characteristics becomes increasingly challenging when the number of finalists for a research project is limited by the nature of data collection (e.g., individual interviews). This presentation describes our use of theoretical sampling (Coyne, 1997) to identify and select a diverse group of research participants by developing and analyzing data from a screening survey (Lavrakas, 2008). Specifically, we implemented a three-stage process to conduct our theoretical sampling. In the first stage, we used our interview protocol to identify the prerequisite knowledge we wanted our participants to have before the interview, and we designed our screening survey questions to assess students' ability to solve problems requiring this knowledge. Second, we developed and used a rubric to evaluate students' thoughtfulness by examining their use of mathematical vocabulary, attention to detail, and the meanings they conveyed in their open-ended survey responses. Finally, we used principles of open coding (Strauss & Corbin, 1998) to identify students who exhibited different meanings of the mathematical concepts in the survey, and, when necessary, incorporated students’ mathematical course experience as an additional criterion to winnow potential participants. Implementing this process ultimately led us to select 6 participants with varying levels of mathematical knowledge who exhibited a wide range of meanings for the topics in our screening survey, which we inferred would yield diverse meanings during our interview tasks. This presentation serves as a practical example for those interested in mathematics education research to utilize in identifying high-quality participants for a small qualitative study.
Nice work, ladies!
