ML Interview Q Series: How does a standard autoencoder contrast with a variational autoencoder, and in which scenarios would one opt to use a VAE instead of a basic autoencoder?