How can I compute the time complexity of the training of an Autoencoder that take in input an array of dimension 1xN and have only one hidden layer with with M neurons. So the matrices of weights are NxM for the hidden layer (encoder weights) and MxN for the output layer (decoder weights).
For both feed-forward and back-propagatation, igmoid activation function is used.
To study the time complexity, do I have to consider also the number of epochs of the training, or the time complexity is referred to only to a trained NN (Autoencoder in my case)?
Thanks a lot!