What is Denoising Autoencoder?
A Denoising Autoencoder is a type of autoencoder designed to solve the problem of having more hidden nodes than input nodes.
Here is a simple structure of a Denoising Autoencoder.
X1 to X4 are the real inputs. Copy those values to use as the input of the Denoising Autoencoder. Then, select a random number of nodes from the input nodes and set them to 0.
When you activate the autoencoder with the newly created input values, the output values are paired with the real input values.
This prevents the autoencoder from simply copying the input to the output.