In neural network toolbox for the function: trainautencoder, how can I use purelin for encodertranferfunction?

I am using the function "trainautencoder" in the neural network toolbox. I can set the decodertransferfunction to be purelin, but the encodertransferfunction cannot be set to purelin. Does anybody know why? And is there a way to set encoder function to also be purelin. My goal is to set up a LINEAR autencoder. Is there a better way then to make a linear atoenncoder?

Answer 1

[Greg Heath]

Using purelin in a hidden layer is a waste of time.

You can get the same result by removing the layer and adjusting the output weights of the previous layer.

Hope this helps.

Thank you for formally accepting my answer

Greg

[david hartman]

But in an autoencoder, since there are two layers- encode and decode, how can one remove the activation of encode layer? It comes with autoencoder.