Professional Documents
Culture Documents
Mathematical Induction: A Systematic Technique in Analyzing Mathematical Proofs
Mathematical Induction: A Systematic Technique in Analyzing Mathematical Proofs
Mathematical Induction: A Systematic Technique in Analyzing Mathematical Proofs
1. Introduction
The principle of mathematical induction has been used for about 350 years. It
was familiar to Fermat, in a disguised form, and the first clear statement seems to
have been made by Pascal in proving results about the arrangement of numbers now
known as Pascal's Triangle. There are many applications of inductive arguments and
the aim of my talk is to give some examples, illustrating why this method has become
an indispensable tool for mathematicians (Bather, 1994).
Principle :
Suppose that (i) p1 is true and that, for n ≥ 1, (ii) pn ⇒ pn+1, then p1 , p2, p3, ... are
all true.
Perhaps the most familiar applications are concerned with proving statements
like the following.
1 + 2 + ... + n + (n + 1) = 1/2(n + 1) + (n + 1)
= 1/2(n + 1)(n + 2)
pn ⇒ pn+1
A slightly harder exercise is to prove that
12 + 22 + . . . n2 = 1/6n(n + 1) (2n + 1)
2. Induction Rules
Peano induction is merely the simplest and best known inductive rule of
inference. Similar structural induction rules are available for every kind of recursively
defined data-structure, e.g. integers, lists, trees, sets, etc. Moreover, it is not
necessary to traverse such data-structures in the obvious, stepwise manner; they can
be traversed using any well-ordering. An extreme example occurs in a standard proof
that the arithmetic mean is greater than or equal to the geometric mean. This uses
an induction rule that traverses the natural numbers by rst going up in multiples of 2
and then lling in the gaps by coming in down in steps of 1. Nor is induction restricted
just to data-structures; it is possible to induce over the control ow of a computer
program or the time steps of a digital circuit (The Automation Proof by Mathematical
Induction, p.5).
2.1. Noetherian Induction
All of these forms of induction are subsumed by a single, general schema of
Noetherian induction:
where ≺is some well-founded relation on the type τ, i.e. ≺ is an irreflective, anti-
symmetric relation and there are no infinite, descending chains, like . . . ≺ an ≺ . . .
≺ a3 ≺ a2 ≺ a1. The data-structure, control ow, time step, etc., over which induction
is to be applied, is represented by the type τ. The inductive proof is formalized in a
many-sorted or many-typed logical system.
Success in proving a conjecture, P , by induction is highly dependent on the
choice of x and ≺. There is an infinite variety of possible types, τ , and for most of
these types, an infinite variety of possible well-orderings, ≺. Thus choosing an
appropriate induction rule to prove a conjecture also introduces an infinite branching
point into the search space. Controlling it, therefore, requires special heuristic
techniques.
3. Conclusion
In this chapter we have researched the systematic techniques in analyzing
mathematical induction. We have seen that mathematical induction is necessary in
proving results or establishing statements for natural numbers and reasoning about
objects, events or procedures containing repetition. It shares with all deductively
valid conclusions the property that it is necessarily true or true in all possible worlds
in which the givens are true.
It have been set that the purpose of mathematical induction is to be able to
prove an equation works for all natural numbers without actually calculating each
one individually.
Bibliography
Bundy A. [1988], The use of explicit plans to guide inductive proofs, in R. Lusk and R.
Overbeek, eds, `9th Conference on Automated Deduction', Springer-Verlag, pp.
111{120. Longer version available from Edinburgh as DAI Research Paper No.
349.
Bundy A., Stevens A., van Harmelen F., Ireland A. and Smaill A. [1993], `Rippling: A
heuristic for guiding inductive proofs', Arti cial Intel ligence 62, 185{253. Also
available from Edinburgh as DAI Research Paper No. 567.
Bundy A., van Harmelen F., Hesketh J. and Smaill A. [1991], `Experiments with proof
plans for induction', Journal of Automated Reasoning 7, 303{324. Earlier version
available from Edinburgh as DAI Research Paper No 413.
Bundy A., van Harmelen F., Hesketh J., Smaill A. and Stevens A. [1989], A rational
reconstruction and extension of recursion analysis, in N. S. Sridharan, ed.,
`Proceedings of the Eleventh International Joint Conference on Arti cial
Intelligence', Morgan Kaufmann, pp. 359{365. Also available from Edinburgh as
DAI Research Paper 419.
Bundy A., van Harmelen F., Horn C. and Smaill A. [1990], The Oyster-Clam system, in
M. E. Stickel, ed., `10th International Conference on Automated Deduction',
Springer-Verlag, pp. 647{648. Lecture Notes in Arti cial Intelligence No. 449. Also
available from Edinburgh as DAI Research Paper 507.
Bundy A., van Harmelen F., Smaill A. and Ireland A. [1990], Extensions to the
rippling-out tactic for guiding inductive proofs, in M. E. Stickel, ed., `10th
International Conference on Automated Deduction', Springer-Verlag, pp.
132{146. Lecture Notes in Arti cial Intelligence No. 449. Also available from
Edinburgh as DAI Research Paper 459.