How to do dimension reduction from a variational autoencoder

Question

I am thinking about a variational autoencoder. As far as I understand it, in the encoding section you compress to a px1 tensor and then you create a $\mu$ and $\sigma$ of dimensions of my choice (though less than $p$). The decoding layer is then the reverse.

But what if I want to do dimensionality reduction using the VAE? Clearly I don't randomly sample from $\mu$ and $\sigma$. Do I just go to the px1 tensor?

Thanks!

Answer 1

I recommend you check this example from the Keras docs and see what you get at the output of the encoder. At the end of the page you can see a representation of the 2 dimensional latent space of the architecture that is defined previously in that article. Plus you can check this article about using VAEs for time series dimensionality reduction and feature extraction.