Self-Supervised Learning in Transformer Models for Birdsong

Learn how self-supervised learning techniques can be applied to transformer models for analyzing and understanding birdsong patterns in this 53-minute conference talk from the Simons Institute's "Decoding Communication in Nonhuman Species IV" workshop. Explore the intersection of machine learning and bioacoustics as Tim Gardener from the University of Oregon demonstrates how modern deep learning architectures can decode complex avian communication systems. Discover the methodologies for training transformer models on birdsong data without explicit labels, understand the challenges of processing temporal acoustic signals from non-human species, and examine how these computational approaches contribute to broader research in animal communication and comparative cognition studies.