Overcomplete Autoencoder uses higher dimension (number of input units) of latent representation than the input.

The opposite is called Undercomplete Autoencoder