Ramadge, PeterHsia, Jennifer2021-08-172026-09-292021-08-172026-09-292021-042021-08-17http://arks.princeton.edu/ark:/88435/dsp01736667621https://theses-dissertations.princeton.edu/handle/88435/dsp01736667621To make generalizable claims about how the human brain functions, it is important to analyze the brain response of multiple subjects as opposed to the response of a single subject. In this study, we want to model and analyze multiple subjects’ fMRI response data to a shared stimulus.We develop two scalable shared response models for modeling multi-subject fMRI responses to natural stimuli (e.g., movie, audiobook). Our models use the notion of balance to build in the intuitive assumption that all healthy, adult human brains function similarly. The main advantage of our models is their competitive accuracy and significant improvement in scalability compared to the benchmark model. High scalability enables our models to handle multi-dataset, multi-subject fMRI data for large-scale analysis.application/pdfenPrinceton University Senior Theses