Comment by strange_things

It's obvious no? For double descent: the network with a billion parameters is so large that it learns to simulate a neural network itself which first uses the test data set to train itself and then produces the final overfitted values??