How good are variational autoencoders at transfer learning?. (arXiv:2304.10767v1 [cs.LG])

Variational autoencoders (VAEs) are used for transfer learning across various
research domains such as music generation or medical image analysis. However,
there is no principled way to assess before transfer which components to
retrain or whether transfer learning is likely to help on a target task. We
propose to explore this question through the lens of representational
similarity. Specifically, using Centred Kernel Alignment (CKA) to evaluate the
similarity of VAEs trained on different datasets, we show that encoders'
representations are …

alignment analysis arxiv cka components datasets discuss good image insights kernel medical music research transfer transfer learning variational autoencoders