I'm a Bit Confused About This, Right?

I'm not so sure why that is the way that you prefer the mathematical properties so much. I think sometimes misapplying this physics approach is a bit dangerous. But my limited experience in trying to avoid over-fitting is never to do less data. In fact, if you try to, you would try to train the model on one part of data and see whether generalized to other parts of data. Yeah. So that's a current stage in IIT from the mathematical viewpoint but then, yeah, so I'm still a bit confused about this.