Abstrakt:
How can machines learn meaningful representations from data without requiring expensive human annotations? This talk explores self-supervised learning—a paradigm shift that enables AI systems to learn from vast amounts of unlabeled data, much like humans learn through observation. We’ll focus on BYOL (Bootstrap Your Own Latent), a method that trains neural networks to discover useful patterns by comparing different views of the same data point. Unlike traditional approaches that require carefully labeled examples, BYOL learns by predicting how the same entity appears under different transformations. Beyond images, we’ll discuss how these ideas extend to diverse domains: analyzing social networks and community structures (BGRL), understanding neural recordings and biological signals (MYOW), and integrating multiple data modalities like text, images, and audio (BEAST). These techniques open possibilities for any field with abundant unlabeled data—from social media analysis and behavioral patterns to scientific discovery. The talk will be accessible to researchers across disciplines and will highlight opportunities for applying these methods to computational social science, network analysis, and other data-rich domains.
