>You can basically do the same thing again with the improved model, and get an even better one...
It does not make much of a difference if the "performance" is limited by the training data.
It does not make much of a difference if the "performance" is limited by the training data.
Great for milking the investors to the last drop...