Embeddings and Prototypes
The discussion revolves around the process of taking embeddings from Resnet 50 and the role of the projection layer. While nonlinear projection heads were explored, the idea of directly passing embeddings to the prototype for online clustering remains untested. Insights into the experimental choices highlight the complexities of model architecture in machine learning.In this clip
From this podcast
Machine Learning Street Talk (MLST)
SWaV: Unsupervised Learning of Visual Features by Contrasting Cluster Assignments (Mathilde Caron)
Related Questions