2

There are always two kinds of induction in general:

  • Assume that the result holds for $n-1$
  • Assume the the result holds for integers $<n$

For the former one, we only need to check one base case; for the latter one, we always need to check two base cases.

I’m learning group theory. Induction is a very useful tool that is always employed. We always prove by induction on group order $|G|$ and assume that results hold for groups of order $<|G|$. But I find that we always check only one base, namely $|G|=1$! Maybe I misunderstood something, but I really don’t remember seeing any proof checking more than one base case.

So what is the point? Could you give me some ideas? Any help would be appreciated!

Possible examples:

  • Which proofs do you have in mind? – lhf Aug 09 '20 at 15:15
  • 7
    There's some confusion in :"for ... strong induction, we always need to check two base cases." In general, strong induction involves no base case. Of course, in particular cases, you might have a proof that works for large $n$ but not for 1 or 2 (or 17) small values of $n$, and then you have to check those values separately, and you might call them base cases, but that's a peculiarity of the particular proof, not of the method of strong induction. – Andreas Blass Aug 09 '20 at 15:17
  • @Ihf https://math.stackexchange.com/questions/560924/index-of-maximal-proper-subgroup-of-a-solvable-group/560934#comment7120986_560934 –  Aug 09 '20 at 15:18
  • 2
    Show that $S\subset \mathbb N$, $1\in S$ and if $k\in S$ for $k<n$, then $n\in S$ implies that $S=\mathbb N$. You can prove it by "weak" induction if you like. Two base cases are not needed. –  Aug 09 '20 at 15:24
  • You don't need to check groups of smaller order because all statements are true for groups of order at most $0$. – David A. Craven Aug 09 '20 at 15:45
  • @AndreasBlass I suppose one could say the same thing about weak induction, right? (And the usual need for a "base case" is just a consequence of the fact that the induction step often fails for $n=1$?) – Ben Blum-Smith Aug 10 '20 at 14:14
  • @BenBlum-Smith Weak induction, as described in the question, needs a base case. Otherwise, you could prove, by weak induction, that all natural numbers $n$ satisfy $n+7=n$: Assume it's true for $n-1$, i.e., $(n-1)+7=n-1$. Add $1$ to both sides to get $n+7=n$, QED. – Andreas Blass Aug 10 '20 at 14:18

2 Answers2

2

Here are some reasons why one might want to check more than one base case:

  • the argument used in the inductive step cannot be applied for small $n$, hence the proof writer checks those cases separately in the base cases
  • the inductive step (proof for $n+1$) explicitly refers to the cases $n$ and $n-1$ (sometimes also $n-2$ etc.)
  • the writer might want to get a better feeling for what needs to be shown, and therefore checks the statement manually for more $n$ than actually needed.

In general, it is therefore a good idea to start (on a separate sheet of paper) with the inductive step. Afterwards, you know which base cases need to be checked.

The same applies for reading a proof: If you wonder why more than one base case was treated, take a look at the inductive step. The answer will most likely be hidden there.

Zuy
  • 4,962
  • 3
    All sheep are pink. To prove this, we start by showing that all sheep are the same color. As a base case, one sheep is one color. For induction, suppose that any group of $k$ sheep are the same color, and consider a group of $k+1$ sheep. Remove one of the sheep---the remainder are the same color, by induction. Remove a second sheep, and replace the first one. This group is also mono-colored. Therefore the entire collection of $k+1$ sheep has only one color. By induction, all sheep are the same color. Regarding the statement of the result: – Xander Henderson Aug 09 '20 at 15:41
  • 1
    https://www.edp24.co.uk/polopoly_fs/1.6167902.1563459288!/image/image.jpg_gen/derivatives/landscape_630/image.jpg – Xander Henderson Aug 09 '20 at 15:41
  • 1
    (This is an example of your second point; the induction here actually requires a second base case, in which there are two sheep). – Xander Henderson Aug 09 '20 at 15:42
  • 2
    One could also say that this is an example of the first point, since the inductive argument cannot be applied when dealing with two sheep. – Zuy Aug 09 '20 at 15:50
  • Indeed. Still, one of my favorite examples. – Xander Henderson Aug 09 '20 at 16:07
1

Here is an induction that requires more than one base case. Say we have two stamps, one 5 cent and the other 3 cents. I claim that any number $n \geq 8$ can be made using just these two stamps. First we have $5+3=8$ and we have $3 + 3+ 3=9$. From these two base cases we can construct the next number by either replacing a $5$ with two $3$s or replace three $3$s with two $5$s. to complete the induction. Note that we do need the two base cases to ensure there will always be enough $5$s or $3$s to make the replacement for the inductive step.

CyclotomicField
  • 11,925
  • 1
  • 14
  • 32