Using VAEs to Learn Latent Variables: Observations on Applications in cryo-EM
This is an incremental study for researchers in computational biology or cryo-EM, focusing on qualitative observations without new methods or benchmarks.
The paper examined the amortization properties of variational autoencoders (VAEs) in biological applications, finding that the encoder qualitatively resembles traditional explicit latent variable representations.
Variational autoencoders (VAEs) are a popular generative model used to approximate distributions. The encoder part of the VAE is used in amortized learning of latent variables, producing a latent representation for data samples. Recently, VAEs have been used to characterize physical and biological systems. In this case study, we qualitatively examine the amortization properties of a VAE used in biological applications. We find that in this application the encoder bears a qualitative resemblance to more traditional explicit representation of latent variables.