Can someone please give an intuitive explanation of the difference between the Iterative Hard Thresholding VS Iterative Soft thresholding algorithm?
And if we can use these algorithms in an autoencoder with a non-linear activation function?
$\begingroup$
$\endgroup$
