I have designed a 1d CNN with 2 main blocks each has 2 layers of 1d CNN and I am using layer norm. But when I am testing my model, I feel like it is bad at generalization. So my guesses are it is because the model cannot pick up major features from my data which causes it overfit my training data.
I have increased the number of epochs and add in more new data to the validation set just to avoid data leakage. The model gets a bit better with increasing in the testing accuracy, but the issue is still there.
Any thoughts on how should I improve my model? Or maybe I think how to let the model to focus on the global features and trend instead of the locals? And I normalized my data over the range of 1 to -1. I do not know if this is the reason that is causing the overfit/bad generalization.
Thanks in advance.