Questions about the learning setting

[HiMarumo]

Thank you for the great work.

When I trained the model on my environment using the train set (0-7, 9, 10), the validation set accuracy was about 5% lower than the provided trained model. Conversely, the accuracy of the train set has increased, so I think it is overfitting. Did you fine-tune the provided trained model on the validation set?

[jbehley]

sorry for the late reply; I think that we did not fine-tune on the validation set. The validation set was used to monitor the training progress and maybe detect overfitting (when the loss/training set metric is still decreasing/increasing and the validation set metrics are not increasing anymore)

hope that clarifies.