I’m just started learning ANN, and I have a question
If we input a set of data to ANN, let’s say set A. And let ANN train. After trained set A, we input another set B. But before we input set B, will the set A removed in order to give space for set B? Like remove old storage for new one.
If this is true, is it possible to build a “re-train” system. Like if I get wrong in a test, I will pick the ones I did wrong and do them again. Can it apply to AI also? After set A is removed, pick some question they did wrong and make set C and input again.
ANN stores only weights. No data is stored within the model. That said, when someoneelse’s private is used to train a model without their consent, it’s still considered as infringement.
Yes, the weights of the model change if you retrain, and the model can forget what it learned from previous tasks! The weights of the model are frozen when the training is finished; if you reopen training, its “learned” state will deviate to a new one!