Among P, NP, NP hard and NP-complete, only the NP-complete set is restricted to decision problems (those that have a binary solution). What is the reason for this? Why not define it simply as the intersection of NP and NP-hard? And this leads to another question - there must be problems that are not necessarily decision problems and also have the property that any problem in NP can be reduced to them in polynomial time. Is this then a set encompassing NP-complete? Is there already a name for this set?
EDIT: Per the comment by Matt and also the post: What are the differences between NP, NP-Complete and NP-Hard?, its seems P and NP are defined only for decision problems. That would resolve this question apart from why they would be defined this way. But, this seems to be in contradiction to the book Introduction to Algorithms by Cormen et.al. In chapter 34, the section titled "NP-completeness and the classes P and NP", they simply say: "P consists of those problems that are solvable in polynomial time". They even say later, "NP-completeness applies directly not to optimization problems, but to decision problems" but say no such thing about P and NP.
The classes P and NP are indeed classes of decision problems. I don’t have my copy of CLRS handy, but if they’re claiming that P and NP are all problems solvable in polynomial time or nondeterministic polynomial time, respectively, they are incorrect.
There are some good theoretical reasons why it’s helpful to define P and NP as sets of decision problems. Working with decision problems makes reducibility much easier (reducing one problem to another doesn’t require transforming output), simplifies the model by eliminating issues of how big the answer to a question must be, makes it easier to define nondeterministic computation, etc.
But none of those are “dealbreakers” in the sense that you can define analogous complexity classes that involve computing functions rather than just coming up with yes or no answers. For example, the classes FP and FNP are the function problem versions of P and NP, and the question of whether FP = FNP is similarly open.