6

I'm reading Recursive Functions at Stanford Encyclopedia of Philosophy (section 1.4). The following paragraph defines function β which is then used to define variant of Ackermann-Péter function:

enter image description here

What exactly is the type of Iter? The base case Iter(ϕ, 0) suggests it's (N → N) → N → (N → N), but the inductive step has type (N → N) → N → N.

Assuming it's the former and Iter just composes ϕ with itself, the next paragraph doesn't make sense to me because Iter should have return type N → N, while β should be N → (N → N), so you cannot compose β with itself.

What am I misunderstanding? Thanks.


Update: I've checked the Wikipedia's article on Ackermann function and it turns out that the correct defintion is as follows:

$$β(0) = S$$ $$β(x + 1) = Iter(β(x))$$

where:

$$Iter(ϕ)(0) = ϕ(1)$$ $$Iter(ϕ)(x + 1) = ϕ(Iter(ϕ)(n))$$

Siegmeyer
  • 273

1 Answers1

6

You're not misunderstanding, there are a few errors in the text. Your first type signature is right, and the "step" would be more clearly written with a $\circ$ rather than brackets, i.e. $\operatorname{iter}(\phi, n+1) = \phi \circ \operatorname{iter}(\phi, n)$

It's clear there's something wrong at the point that they have an equation with an x on one side but not the other. It should be $\operatorname{iter}(\phi, n)(x) = \phi^{n+1}(x)$

Similarly the definition of $\beta(0)$ is confusing at best. $\beta(0)(x) = x+1$ would be correct, or $\beta(0) = \lambda x.x + 1$

I'd try another source if I were you.

Matthew Towers
  • 11,540
  • 1
    Yes, I normally really like SEP articles but this isn't good. (And btw, to the OP: this is a great example of how $\lambda$-notation is useful, since it makes it very clear what type of thing $\beta(0)$ is.) – Noah Schweber May 25 '20 at 19:54