Advertisement Jump to content
Sign in to follow this  
johnnyBravo

Automating the learning rates for a feedforward backpropagation network?

This topic is 4902 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

Hi, I was thinking of making the learning rate for my feedforward backpropagation network automated. So all you have to do is enter the inputs/desired outputs. What automatic methods of determining the learningRate do you suggest? One idea I read on the internet suggests checking if the new error exceeds the old error by more than a predefined ratio, so if it does the new weights and biases are discarded and the learning rate is decreased by multiplying it against a lrDecrease variable. Otherwise you just increase the learningRate by multiplying against a lrIncrease variable. So how would I go and have the predefined ratios, lrDecrease variables, lrIncrease variables automatically set? Would these variables some how be determined by the data I want?, eg as i am using the sigmoid function and get outputs between 1 and -1, say i wanted accuracy up to 3 decimal places, would I set the lrIncrease to something like 0.0001? Thanks [Edited by - johnnyBravo on August 21, 2005 8:18:45 PM]

Share this post


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

  • Advertisement
×

Important Information

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

GameDev.net is your game development community. Create an account for your GameDev Portfolio and participate in the largest developer community in the games industry.

Sign me up!