How can I choose the best neural network in MATLAB when my results change each time I retrain?

2 views (last 30 days)
Hi. I have finally built a dataset and trained it (for classification) in both patternnet and the Classification Learner App. I usually get an accuracy of over 98%, but each time I retrain the network, this number changes. How do I know which network is the best when each of them vary around 1% each time I retrain? Thanks

Answers (1)

KJVKU999
KJVKU999 on 15 Jun 2017
Edited: KJVKU999 on 15 Jun 2017
You can use at the beginning of your program:
rng(1);
so that your results won't differ every time you retrain your network.
This link will provide more information: https://www.mathworks.com/help/matlab/ref/rng.html

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!