MATLAB Answers

How I can improve the accuracy of over fitted CNN model

3 views (last 30 days)
CHHAVI
CHHAVI on 9 May 2021
Answered: Vineet Joshi on 17 May 2021
I reshaped my data to 4D (9x2000x10x20000) i.e height x width x channel x instances. and label 20000x1. but i am only getting 20% caccuracy. and model is overfitting.

Answers (1)

Vineet Joshi
Vineet Joshi on 17 May 2021
Hi
Overfitting is when the model performs well on training data but not on validation data.
We can see from the provided figure that the model is not performing well on the training data itself, which is unlikely due to overfitting.
Based on your training statistics it also looks like you haven’t even completed a single epoch, which may also be the cause of underperforming model as larger datasets often requires multiple epochs.
You can wait for the training to finish and based on the achieved accuracy, try experimenting with different architectures and hyperparameters according to your problem. This may help you in increasing the training accuracy.
The following link can direct you to many resources for CNNs with MATLAB.
Hope this helps.
Thanks

Community Treasure Hunt

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

Start Hunting!