You are cordially invited to a CCN Seminar with John Pearson, Assistant Professor of Biostatistics & Bioinformatics, Duke University.
To set up a meeting with John while he is here, please reach out to Jessica Hauser at email@example.com.
Title: "In improvisation, there are no mistakes": joint modeling, variability, and learning in birdsong
Abstract: As the last decade has witnessed an explosion in the volume and complexity of neuroscience data, finding methods that successfully link high-dimensional brain activity to complex behavior has emerged as one of the key challenges in statistical neuroscience. In particular, such joint models should flexibly handle both missing data and data of disparate dimensions, as well as providing reproducible results. To this end, I will discuss recent work from my group on joint modeling in the context of zebra finch song copying, a complex vocal-motor learning problem in which juveniles must first memorize and then reproduce the song of an adult tutor--without the benefit of external reinforcement. Here, variational autoencoders allow us to quantify the multidimensionaly variability in vocal performance, track its structure over the course of learning, and link these variations to patterns of local activity in song-specific regions of the basal ganglia. Finally, I will discuss work in progress on identifiable variational autoencoders for the joint modeling problem that allow us to view the relationships between data modalities as arising from interactions between separate latent spaces