python - Validation loss way greater than 1

Question

I'm using Keras to classify EEG signals, which are processed with different feature extraction methods. I constantly run into an issue, where the Validation loss is way greater than 1 regardless of the method used. I am using Sparse Categorical Crossentropy as a loss function, and the ANN is a traditional resolution neural network with 50 layers (ResNet50). The validation set is made by choosing random samples from the training set. Hope someone has an idea and I really hope it is caused by a silly mistake.

question from:https://stackoverflow.com/questions/65833383/validation-loss-way-greater-than-1

Answer 1

Waitting for answers