0
$\begingroup$

I want to train both a single-layer autoencoder and a multi-layer autoencdoer in Keras to reconstruct an input with 24 features, all in the same scale with int values from 0 to ~200000. My question is: what would be the best choice for activation function for each layer for both autoencoders?

In the Keras autoencoder blog post, Relu is used for the hidden layer and sigmoid for the output layer. But using Relu on my input would be the same as using a linear function, which would just approximate PCA. So what would be a better choice to learn non linear features?

Same goes for the multi-layer autoencoder. In the blog post only Relu is used for the hidden layers and sigmoid for the output layer. Do better options for my training data exist?

I plan to use MSE as loss function. Also, Is it necessary to scale/normalize my input data given that all features are already in the same range? And in that case, which scaling would be best?

$\endgroup$
1
$\begingroup$

Since the activation is applied not directly on the input layer, but after the first linear transformation -- that is, $\text{relu}(Wx)$ instead of $W\cdot \text{relu}(x)$, relu will give you the nonlinearities you want.

And it makes sense for the final activation to be relu too in this case, because you are autoencoding strictly positive values.

Scaling and normalization is still important, because the initialization of neural network weights is carefully chosen so that for reasonably scaled inputs, the optimization process is greatly eased. A simple scaling of the inputs to around [0,1] should do the trick.

$\endgroup$
  • 1
    $\begingroup$ I tried several parameters and, almost independently of the scaler used (including no scaler), a linear activation in the output layer works almost always better than relu. Which I find odd. $\endgroup$ – RiccB Mar 26 '18 at 10:38
  • $\begingroup$ It's possible that using linear instead of relu at the end helps the gradients flow better and avoids the dying relu problem, so I wouldn't be too surprised. $\endgroup$ – shimao Mar 26 '18 at 10:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.