Special numeric values
In addition to dtype objects, NumPy introduces special numeric values: nan and inf. These can arise in mathematical computations. Not A Number (nan). It indicates that a value that should be numeric is, in fact, not defined mathematically. For example, 0/0 yields nan. Sometimes, nan is also used to signify missing information; for example, pandas uses this. inf indicates a quantity that is arbitrarily large, so in practice, it means larger than any number the computer can conceive of. -inf is also defined and it means arbitrarily small. This could occur if a numeric operation blows up, that is, grows rapidly without bounds.
Nothing is ever equal to nan; it makes no sense for something undefined to be equal to something else. You need to use the NumPy function isnan to identify nan. While the == sign does not work for nan, it does work for inf. That said, you're better off distinguishing finite and infinite values using the function is finite or is inf. Arithmetic involving...
