Instance-Wise Denoising Autoencoder for High Dimensional Data

Algorithm 1

Instance-Wise Denoising Autoencoder (IDA).

Input: Training data , zeros probability in mask vectors

While   (error threshold) do

  for    to    do

     Generate random mask boolean vector according to the density :

       where

     Perform denoising with mask:

   Get corrupted inputs

   Compute autoencoder according to different , , , :

  end

  Obtain the final solution by combining autoencoders together:

end

Return: