I will discuss some stylized topological inference problems and give some information about Kolmogorov, sample and computational complexity. I will then discuss some tentative first steps towards a theory of feasibly computable invariants.