Jump to content
  • Advertisement
Sign in to follow this  

Whats the minimum training error you expect to achieve?

This topic is 4349 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

Lets say you are training a neural network to learn a function. The training set is 1000 point long and has been normalized to lie between 0.1-0.9. What is the training MSE that you would find acceptable, where MSE is calculated as (Sum_over_all_points((value_actual - value_predicted)^2 ))/total_number_points total_number_points=1000=length of trainin set thanks a lot.... PS: my current values are something like 4e-3, and that doesnt seem too good!!

Share this post


Link to post
Share on other sites
Advertisement
A good way to test your NN is to have another set for test purposes. Once you have trained your net with trainning set, see the error with the test set.

If the errors are very different, probably you have overtrained your net, so instead of solving the problem, it has memorized all the trainning set. If you get a very small error, that´s probably what´s happening to you.

Greetings!

Vicente

Share this post


Link to post
Share on other sites
Sign in to follow this  

  • Advertisement
×

Important Information

By using GameDev.net, you agree to our community Guidelines, Terms of Use, and Privacy Policy.

Participate in the game development conversation and more when you create an account on GameDev.net!

Sign me up!