For example for PCA, the proportion of variance explained is proportional to the eigenvalue of the respective feature.
Now for VAEs, is there a way to estimate the amount of variance that is explained by a single latent factor of variation?
Or assume given a (synthetic) image data sets, potentially including a lot of noise, where some of the data generating factors are labeled. Is there a way to estimate the amount of variance that is explained by these labels?
This is especially challenging as the underlying processes are most likely highly non-linear.
