Hi Folks,
I am working on my thesis and using the C-MAPSS dataset. There, I have been using Transformer architecture to synthesize the data. Here, I use the first 80 engine IDs for training purposes and the remaining 81-90 engine IDs for validation purposes, and thereafter, I use 91-100 engine IDs for test purposes. I did not use train_test_split functionality to get train and validation datasets. I have also created two extra attributes in the data frame to classify binary and multiclass classification. However, after training, my training loss is greater than my validation loss. Is there any way can resolve this issue?