DAE, robust autoassociative network
Improved the robustness of representations
It intentionally corrupting the inputs of a standard autoencoder during training. (Gaussian Noise)
The problem of predicting the original input from an input mixed with noise.