cnn validation accuracy not increasing

126 views (last 30 days)
new_user
new_user on 20 Dec 2021
Edited: Prince Kumar on 6 Apr 2022
I am not able to increase validation accuracy after 70s. The traing curve is not too smooth. The loss is not saturating! the 2nd image 1001.png has increasing loss after some iteration. I am using laky relu, what to do to reduce the error, saturate the loss curve and increase accuracy. What to do, any suggestions highy appreciated.

Answers (1)

Prince Kumar
Prince Kumar on 6 Apr 2022
Edited: Prince Kumar on 6 Apr 2022
Hi,
According to the attached screenshot, the model is overfitting. This generally happens when your model is learning the data instead of learning the pattern.
Following few thing can be trieds:
  • Lower the learning rate
  • Use of regularization technique
  • Make sure each set (train, validation and test) has sufficient samples like 60%, 20%, 20% or 70%, 15%, 15% split for training, validation and test sets respectively.
  • Perform k-fold cross validation
  • Randomly shuffle the data before doing the spit, this will make sure that data distribution is nearly the same. If your data is in datastore you can use 'shuffle' function else you can use "randperm" function.
Hope this helps!

Community Treasure Hunt

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

Start Hunting!