I have a CNN model that classifies 10 classes of audio spectrograms. However, since I work with the open set of data, I need to classify the unknown audio data as an "Unknown" class. The problem is my training samples of unknown data are larger than the other known class. I'm afraid that there would be a problem when the model performs stochastic optimization.
Should I separate the "Unknown" training data and train the model separately. Or I can just simply mix the unknown data to the other classes and train the model right away?