I'm doing a project that uses a clustering algorithm for the facial expression classification task. So, I use the output of the encoder in the VAE autoencoder for dimensionality reduction. However, I learned that they use a random sample noise based on a normal distribution of the output's encoder.
So, I'm confused about whether I should you the output of the encoder after noise sampling or any layers in the VAE.
Also, do you know any better autoencoder architecture for dimensionality reduction?