Is this Quantum Neural Network overfitting?

PennyLane Help
1

In the accuracy graphs (attached the graph images below) shown in this code (Binary Classification for Fraud Detection):

  1. validation loss is greater than training loss
  2. training accuracy is greater than validation accuracy

Does that not mean that the model is overfitting? How to avoid that overfitting?

image
image722×585 42.5 KB

image
image707×572 46.4 KB

2

Hi @ilkayn, great question!

You’re correct in your analysis that conditions 1 and 2 can indicate overfitting. In this case however the difference in loss and accuracy between training and validation is so small that it shouldn’t be a big concern.

In more general terms, if you get to a situation when you’re experiencing overfitting, one path to take is to reduce the number of parameters in your model. This will produce training results that are not as good but it can improve your validation results.

Please let me know if this answers your question!

1 Like
3

Ok, will try that, thank you!