7

I came up with this:

(S.1) $a - a = 0$

(S.2) $a - b = S(a - S(b))$

This seems to work. At least for $a$$\ge$$b$.

Is this the correct or most efficient formulation?

Also, does there happen to be one for division? Of course I would imagine that that division would only work for $a/b$, where $a$ is a multiple of $b$.

What I've come up for this one is:

(D.1) $0 / a = 0$

(D.2) $a / b = 1 + (a - b)/b$

David Bandel
  • 333

3 Answers3

7

From an implementation point of view, your definition is inefficient, because at each step you have to to decide whether to apply the $a-a = 0$ rule, which requires that you check to see if the two arguments of $-$ are equal, and this will take a long time if the arguments are large.

A more efficient implementation is:

$$\begin{array}{rll} a&-0 &= a \\ 0&-S(b) &= 0 \qquad\text{(or leave this undefined)}\\ S(a)&-S(b) & = a-b \end{array}$$

Here you can decide in constant time which of the three cases applies: is the second argument zero? If so apply the first rule, if not, is the first argument zero? If it is apply the second rule; if not apply the third rule.

Leave the middle case undefined if you want ordinary subtraction, but defined as 0 if you want so-called "truncated subtraction" in which $a-b = 0$ when $a<b$.

For division, you are better off splitting it into integer operations. For every $a$ and $b$ there is an integer quotient $q$ and an integer remainder $r$ such that $$a = qb+r$$ and $0\le r < b$; if $r = 0$ then $b$ divides $a$ exactly and $q = a\div b$. Calculating the integer quotient and remainder in Peano arithmetic isn't hard. The quotient is:

$$\begin{array}{rll} 0&\div S(b) &= 0 \\ S(a)&\div S(b) &= S((a∸ b)\div S(b)) \end{array}$$

Where ∸ denotes truncated subtraction. Then the remainder is just $a - b\cdot(a\div b)$.

MJD
  • 65,394
  • 39
  • 298
  • 580
  • The division algorithm I suggested doesn't work. I will try to correct it, but not today. – MJD Nov 11 '16 at 03:37
  • The division algorithm you suggested seems to work at every step except the final step where you perform 0-a=0 (truncated subtraction). As such the result will always be 1 higher than the actual quotient. Any ideas on how to fix this? – Ozaner Hansha May 24 '18 at 20:09
  • I don't see it, and the three examples I tried seem to work. Could you please give an example? – MJD May 25 '18 at 03:56
  • 5/3 = S(4)/S(2) = S((4-2)/S(2)) = S(2/S(2)) = S(S(1)/S(2)) = S(S((1-2)/S(2))) = S(S(0/S(2))) = S(S(0)) = 2 – Ozaner Hansha May 25 '18 at 13:15
  • Looking at it again, I think it only fails if $a$ doesn't divide $b$ evenly. In this case it always seems to be 1 higher. – Ozaner Hansha May 25 '18 at 14:15
  • 2
    @MJD I think $S(a) \div S(b) = S((S(a) - S(b)) \div S(b))$ is the correct formula. – corvus_192 Jul 13 '18 at 12:28
  • How can we define subtraction as a function symbol without presupposing it in the definiendum? – Cherry Blossom Bomb Mar 17 '21 at 06:18
1

Your definition would work, and as Peano only defines the natural numbers, you would only need subtraction when $a \geq b$. Normally the Peano axioms do not define subtraction, but subtraction is just defined as the inverse of addition, i.e. $a - b = a + (-b)$, where $-b$ is the number defined by $b + (-b) = 0$. For this to work you need the whole $\mathbb{Z}$ to work with though.

Greebo
  • 496
0

If all you want is a definition, the simplest one should be: $$a-b = c \iff a = b+c$$

If you want a constructive formula, you want the $S$ on the left. I'd suggest:

  • $a-a = 0$ (same as your S.1)
  • $S(a) - b = S(a-b)$

This of course only works if $a\ge b$, but if $a<b$, the difference is not defined in the natural numbers.

celtschk
  • 43,384
  • 1
    I'm not sure I understand how S(a) - b = S(a - b) works.

    Example: 4 - 1 becomes S(3) - 1 = S(4 - 1), which doesn't seem to take us anywhere.

    Whereas, using my definition 4 - 1 becomes S(4 - S(1)) = S(4 - 2), which eventually will recurse down to S(S(S(0))) which evaluates to 3. I may be misunderstanding what you mean by "constructive."

     – David Bandel Aug 07 '13 at 08:32
  • 4 - 1 becomes S(3) - 1 = S(3 - 1), so now you have to compute 3 - 1 instead of 4 - 1. Then you work your way down to S(S(S(1-1) = 3 – Greebo Aug 07 '13 at 08:51
  • I see. But how is this any better or more "constructive" than my formulation? – David Bandel Aug 07 '13 at 08:55