0
$\begingroup$

Can somebody please explain how to train a neural network in batch mode. I have a single target time series of length $N$ for a given input time series of the same length. In order to apply Hopfield network's content addressable property, I want to store this pattern of input time series. Then given a test time series that is a noisy version, how do I apply Hopfield neural network with Particle Swarm or Genetic algorithm as a learning method with the MSE as the fitness function? All this is confusing, since I am well practiced with incremental training and not batch wise training. PSO or GA update their equations for every input, so how do I adapt this with batch wise training. Insights on this technique will help to clear away the confusions. Thank you

$\endgroup$
1
$\begingroup$

Use the whole data set for one increment (which you claim to understand). Its now a batch update. It will take many updates to converge.

$\endgroup$
1
  • 1
    $\begingroup$ Then it sounds more like you are in way over your head. I would recommend learning more about the individual parts. Trying to teach you how each of those work would require weeks of lecture and self study, and is in no way appropriate for a QA format. $\endgroup$ Mar 26 '14 at 5:00

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.