General question on generalization bounds in Q2-3

[melipone]

What are those different bounds good for? Isn't there just one learning problem? Why would we choose one over another?

[yaser]

Originally Posted by melipone

What are those different bounds good for? Isn't there just one learning problem? Why would we choose one over another?

These are theoretical attempts to bound the generalization error using different approaches. Some of the bounds are more successful than others (read: are closer to the actual value of the generalization error) in different ranges. The goal of the problems is to compare them.