Is there a natural topology on the set of open sets?

In machine learning, is it better to train your model iteratively across several training/validation sets or is it better to just include all the data in your training set the first time and feed that through the model and why?

  • By "train your model iteratively across several training/validation sets," I mean split original training set into several data sets, train on remaining training records, predict on validation set 1, append the predicted values to validation set 1, train on validation set 1, predict on validation set 2, append the predicted values to validation set 2, train on validation set 2, predict on validation set 3, etc.; the logic being that the model would learn from the previous mistakes every iteration.

  • Answer:

    I'm not sure how you would be able to determine how good your model was from this.  Certainly you would get different results based upon how your ordered your validation sets.

Ralph Winters at Quora Visit the source

Was this solution helpful to you?

Other answers

it depends on the algorithm for an traditional SVM classifier, this would have no effect this is true of any convex method, up to round off / numerical errors what you are describing is similar to a  transductive SVM and this might affect the convergence of this or any non-convex method (deep learning / neural network  as well )

Charles H Martin

I can't speak for everyone, but this is how the process goes for good NLP work: Have three sets of data: A training set - make your model A development set - tune your model A test set - try your model on it. DONT go back to the first step once you've tried it on your test set or you risk significant bias. These data sets should be distinct and as large as you can make them. There are methods in ML used in NLP that use iterative methods in developing the model (MCMC), but you should still have an independent test set to try your model on to see if it's good once it's developed.

Anonymous

Find solution

For every problem there is a solution! Proved by Solucija.

  • Got an issue and looking for advice?

  • Ask Solucija to search every corner of the Web for help.

  • Get workable solutions and helpful tips in a moment.

Just ask Solucija about an issue you face and immediately get a list of ready solutions, answers and tips from other Internet users. We always provide the most suitable and complete answer to your question at the top, along with a few good alternatives below.