Machine Learning Street Talk (MLST) cover image

SWaV: Unsupervised Learning of Visual Features by Contrasting Cluster Assignments (Mathilde Caron)

Machine Learning Street Talk (MLST)

00:00

Optimizing SWaV: Clustering and Learning Techniques

This chapter explores optimization techniques related to the SWaV method, focusing on adapting learning schedules and batch sizes for effective training. The discussion emphasizes the complexities of clustering in self-supervised learning, including the balance between prototypes and assignments, and how unsupervised methods can sometimes outperform supervised learning. It also highlights the importance of general-purpose representations and concludes with insights on the future of visual feature learning.

Transcript
Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app