Register to reply

Cantor's Diagonalization Proof of the uncountability of the real numbers

Share this thread:
micromass
#55
Feb6-12, 08:52 PM
Mentor
micromass's Avatar
P: 18,346
We all heard such quotes before. Earlier they might have said:

"Negative numbers is a disease from with future mathematics will someday be cured"

or

"Imaginary numbers is a disease from with future mathematics will someday be cured"

or

"Noneuclidean geometry is a disease from with future mathematics will someday be cured"

or

"Decimal notation is a disease from with future mathematics will someday be cured"

These statements are all quite incorrect. Negative numbers, imaginary numbers, noneuclidean geomety and transfinite numbers are ALL very useful concepts. They will NOT go away.

Transfinite numbers yield a perfectly consistent theory (in so far that the natural numbers are consistent), and they form a theory that are very applicable. They show up literally everywhere in mathematics.

Cantor's ideas were initially despides and Cantor ended up in a deep depression. People like Kronecker and Poincarre were (although they were very smart) ultimately conservative and did not foresee the huge successes in axiomatic set theory.

I understand you don't like it. But you have to realize that it will not go away because mathematicians like it. Transfinite numbers might be unintuitive, but they eventually are entirely logically sound.

I notice a trend. Initially an innovative new idea is always laughed at, then it is furiously attecked and in the end it is accepted as common sense. (somebody else said this, but I can't remember who). Right now, transfinite numbers ARE a common sense. Mathematics will not be cured from axiomatic set theory, and if they would remove axiomatic set theory from mathematics then they would rip out its very heart.
pwsnafu
#56
Feb6-12, 09:06 PM
Sci Advisor
P: 838
Quote Quote by Leucippus View Post
Henri Poincare would be far more approving of infinity thought of as an endless process.
In that case Henri Poincare should not work on Wall Street.
andrewr
#57
Feb6-12, 09:10 PM
P: 263
Quote Quote by MrAnchovy View Post
I am a mathematician (of sorts), but I am trying to present an argument to someone who isn't in terms that he seems to understand.
I have this argument over aleph not occasionally with my brother who is a Mathematician, and he and I both have a problem with the "Axiom of Choice" as he calls it.
Let me ask a simple question, If I were to show you a way to count every possibly decimal representation as a unique list -- and that means as many digits as can be imagined (inductively given n digits, I will list n+1 as well...) What would it do to this silly argument?

I realize many people think it is impossible; but that is part of the point.
I believe it can be done -- is it sufficient to put a big !"?"! into the whole idea of Cantor's idea?

(I am serious... and this isn't a difficult proof.)
micromass
#58
Feb6-12, 09:15 PM
Mentor
micromass's Avatar
P: 18,346
Quote Quote by andrewr View Post
I have this argument over aleph not occasionally with my brother who is a Mathematician, and he and I both have a problem with the "Axiom of Choice" as he calls it.
Let me ask a simple question, If I were to show you a way to count every possibly decimal representation as a unique list -- and that means as many digits as can be imagined (inductively given n digits, I will list n+1 as well...) What would it do to this silly argument?

I realize many people think it is impossible; but that is part of the point.
I believe it can be done -- is it sufficient to put a big !"?"! into the whole idea of Cantor's idea?

(I am serious... and this isn't a difficult proof.)
I really don't understand what you're trying to say. The axiom of choice has nothing, absolutely nothing to do with Cantor's diagonalization.
Deveno
#59
Feb6-12, 09:38 PM
Sci Advisor
P: 906
Leucippus, let me offer you a variant of Cantor's proof, that show that your consideration of "slope" is irrelevant:

we proceed as before, except, for our "diagonal number" we do THIS:

if the first digit of the first number is 1, we assign the diagonal number the first digit 2.

otherwise, we assign the first digit of the diagonal number to be 1.

the next 8 digits of the diagonal number shall be 1, regardless.

if the 10th digit of the second number is 1, we assign the diagonal number the 10th digit 2.

otherwise, we assign the number 1 to the 10th digit of the diagonal number.

the next 89 digits of the diagonal number are assigned the digit 1.

if the 100th digit of the third number is 1, we assign the digit 2 to the 100th digit of our diagonal number.

otherwise, we assign 1 to the 100th digit of our diagonal number.

we continue in this way.

now the "slope" is "even worse" the diagonal is not even as steep as before. nevertheless, no matter which n we pick, the diagonal number differs from the first n numbers in our list, so it differs from every number on our list (it differs from the k-th number in the 10k-1-th digit).

@ micromass: i disagree with the contention that if a proof can only be demonstrated graphically, it isn't valid. for what is a formal system, if not a few squiggles on paper? is that not graphical? i don't know about you, but i read most of my mathematics.

more to the point: many proofs in algebraic topology or category theory amount to: diagram-chasing. the diagram IS the proof, in some cases. there's no logical reason why "pictures" can't be made every bit as unambiguous and consistent as well-formed formulas. constructing a length with a compass and straight-edge IS a proof of constructibility (a concept which can be expressed purely algebraically).

there is many people who think that a mathematical result "isn't true" until it is proven formalizable in (first- or second-) order logic in a formal language that has encoded the axioms of Zermelo-Frankel set theory. this is unfortunate.
MrAnchovy
#60
Feb6-12, 09:41 PM
P: 542
Quote Quote by andrewr View Post
If I were to show you a way to count every possibly decimal representation as a unique list -- and that means as many digits as can be imagined (inductively given n digits, I will list n+1 as well...) What would it do to this silly argument?
Any countable list can be used as the start of Cantor's argument: simply construct a number whose nth digit is different from the nth number on the list for every n and you have a number that is not on the list.
Jorriss
#61
Feb6-12, 09:49 PM
P: 1,072
This is kind of off topic but I'm curious why graphs can not be proofs (I have not taken a course on logic)?
micromass
#62
Feb6-12, 09:52 PM
Mentor
micromass's Avatar
P: 18,346
Quote Quote by Deveno View Post
more to the point: many proofs in algebraic topology or category theory amount to: diagram-chasing. the diagram IS the proof, in some cases.
A commutative diagram is not a proof. And many mistakes in category theory actually amount to misleading diagrams. A commutative diagram is a suggestion of a proof. It is merely an illustration.

If you look at Bourbaki, then they show that they treat commutative diagrams very rigorously. They try to prove everything with algebraic means.

To me, a proof is correct if it formilizable in some kind of logical system. You don't actually NEED to formalize it, but just giving a suggestion that it can be done is good enough.
Hurkyl
#63
Feb6-12, 10:01 PM
Emeritus
Sci Advisor
PF Gold
Hurkyl's Avatar
P: 16,091
Quote Quote by Leucippus View Post
Precisely what Henri Poincare warned Cantor not to do (i.e. not to think of infinity as a completed thing)
Ah joy, "completed infinity".

Just for your information, I have never encountered somebody who uses that terms that could give an even quasi-rigorous account of what the term means.

From observing people who use it, I quite honestly believe that it literally means "I am unwilling or unable to engage in one more layer of abstraction, and thus have failed to recognize that we're talking about the same idea." (I don't mean insult by this, except possibly in the 'unwilling' case)
micromass
#64
Feb6-12, 10:02 PM
Mentor
micromass's Avatar
P: 18,346
Quote Quote by Jorriss View Post
This is kind of off topic but I'm curious why graphs can not be proofs (I have not taken a course on logic)?
Hmmm, I don't think I'm expressing myself clearly enough.

There are actually two kind of proofs: informal and formal proofs.

By the very definition, a proof is formal. A proof is defined as a sequence of logical algebraic steps. Each step follows from another step by applying an inference rule.

These things are the actual proofs. However, they are EXTREMELY hard to read. If all math books would consist out of formal proofs, then math books would be unreadable.

That's why we allow informal proofs. An informal proof presents people the basic ideas and reasoning of a proof. An informal proof suggests that the formal proof can be carried out if one would want to do so.

All text books contain informal proofs (and it's good that they do). Text books like to work with a lot of text and a lot of pictures to make it easier on the student. But one should always keep in mind that one should be able to formalize the proof.

And indeed, one can formalize all the proofs. An example is given by http://us.metamath.org/index.html which contains thousands of proofs including the celebrated Hahn-Banach theorem. This site shows what actual formal proofs look like. And as one can see, it is completely unreadable. (the site also shows formal proofs in plane geometry, including constructibility).

I am NOT advocating that everybody should be using formal proofs in journals or education. Quite the contrary. But the students should be well-aware that the formal proof can be carried it if one wants to do it.
Jorriss
#65
Feb6-12, 10:05 PM
P: 1,072
Quote Quote by micromass View Post
Hmmm, I don't think I'm expressing myself clearly enough.

There are actually two kind of proofs: informal and formal proofs.

By the very definition, a proof is formal. A proof is defined as a sequence of logical algebraic steps. Each step follows from another step by applying an inference rule.

These things are the actual proofs. However, they are EXTREMELY hard to read. If all math books would consist out of formal proofs, then math books would be unreadable.

That's why we allow informal proofs. An informal proof presents people the basic ideas and reasoning of a proof. An informal proof suggests that the formal proof can be carried out if one would want to do so.

All text books contain informal proofs (and it's good that they do). Text books like to work with a lot of text and a lot of pictures to make it easier on the student. But one should always keep in mind that one should be able to formalize the proof.

And indeed, one can formalize all the proofs. An example is given by http://us.metamath.org/index.html which contains thousands of proofs including the celebrated Hahn-Banach theorem. This site shows what actual formal proofs look like. And as one can see, it is completely unreadable. (the site also shows formal proofs in plane geometry, including constructibility).

I am NOT advocating that everybody should be using formal proofs in journals or education. Quite the contrary. But the students should be well-aware that the formal proof can be carried it if one wants to do it.
I better see what you're getting at now. Thank you for clarifying.
micromass
#66
Feb6-12, 10:08 PM
Mentor
micromass's Avatar
P: 18,346
Quote Quote by Jorriss View Post
I better see what you're getting at now. Thank you for clarifying.
Just to clarify it some more: if I see a graph somewhere which intuitively shows me something, then I will accept this as proof. Any mathematical journal (except logic-journals) will accept it as a valid proof. If you write it on the exam, then the professor will accept it as proof (except in a logic class). But it's not a formal proof in the strict sense of the word.
Hurkyl
#67
Feb6-12, 10:14 PM
Emeritus
Sci Advisor
PF Gold
Hurkyl's Avatar
P: 16,091
mm: I think the point Deveno makes (and maybe that Jorriss is wondering about) is that proof theory can be cast in settings other than string processing.

For example, arithmetic expressions and equations in category theory are often presented in graphs.

In Categories, Allegories, Freyd introduces the idea of a Q-sequence (and a Q-tree) that consists of a path through a category and a sequence of quantifiers, which can replace the role of formulas.

Applied to the category of finitely-presented categories, he uses them to prove a metatheorem that an equivalence functor preserves and reflects a property if and only if it can be expressed on a blackboard.
andrewr
#68
Feb6-12, 10:25 PM
P: 263
Quote Quote by micromass View Post
I really don't understand what you're trying to say. The axiom of choice has nothing, absolutely nothing to do with Cantor's diagonalization.
IN my discussions of Aleph Not with my brother, the Axiom of choice does come up.
But, set that aside -- what I am saying is that I believe I can show an algorithm which will demonstratably count every possible decimal number in the decimal notation system, uniquely, and once.

eg: 1.01 will be counted exactly once, as will 2.01, 2.011, 2,0000001, etc.
I am not trained in formal mathematical proofs with respect to Aleph not, etc, so excuse my vagueness.
However, my understanding is that the Real numbers are representable by the decimal notation system; and that the "Ideal" representation of all real numbers is spanned by a mantissa and fraction portion that becomes more and more perfect as the number of digits in each is allowed to grow arbitrarily large.

If I were to demonstrate an algorithm which showed that all possible decimal numbers can be counted with a distinct index; say 'i' so that any imaginable real number would exist "somewhere" in a sequence defined by an index (i) which is an integer from 0 to as large as one likes. eg: every possible decimal number could be found by a static algorithm called f(i).

The only issue which would come up is the representation of 0.99999999.... vs. 1.00000.... and whether or not they are distinct numbers as i becomes arbitrarily large.

If I am following the argument correctly, Cantor appears to be claiming that such a unique counting is impossible. So if I gave a simple counter-example in decimal notation -- would such an algorithm be of use in this discussion?
Hurkyl
#69
Feb6-12, 10:47 PM
Emeritus
Sci Advisor
PF Gold
Hurkyl's Avatar
P: 16,091
Quote Quote by andrewr View Post
eg: 1.01 will be counted exactly once, as will 2.01, 2.011, 2,0000001, etc.
If I'm understanding what you describe, then you truly do generate a list containing every finite decimal, no matter how long. The set of terminating decimals is, indeed, countable.

The problem is that most real numbers don't have finite decimal representations. It sounds like you have a specific algorithm in mind; can you write down the first 10 or so numbers? After that, can you tell me at what place the number 1/9 appears?
andrewr
#70
Feb6-12, 11:00 PM
P: 263
Quote Quote by Hurkyl View Post
If I'm understanding what you describe, then you truly do generate a list containing every finite decimal, no matter how long. The set of terminating decimals is, indeed, countable.

The problem is that most real numbers don't have finite decimal representations. It sounds like you have a specific algorithm in mind; can you write down the first 10 or so numbers? After that, can you tell me at what place the number 1/9 appears?
Yes, and Yes, although your second question can't be written down inside a text terminal of this size. Representation of an integer is not the same as existence of that same integer.
There are many counting numbers which we know exist, but can't be written down inside a text terminal of this size.

You would agree, no doubt, that "0.999999999....." is a representation of a real number; and by the ellipsis I am suggesting a pattern for something which is much larger than I would like to write down, although I DO KNOW HOW to write it down -- eg: and that the representation improves in accuracy and may have a selectable accuracy for a given discussion or purpose.

The algorithm itself can be adapted for various constraints; eg: there are many such algorithms based on a similar methodology.

DO you care if I count only the positive reals, or do I need to include the negative as well?
etc.
Deveno
#71
Feb6-12, 11:14 PM
Sci Advisor
P: 906
Quote Quote by andrewr View Post
Yes, and Yes, although your second question can't be written down inside a text terminal of this size. Representation of an integer is not the same as existence of that same integer.
There are many counting numbers which we know exist, but can't be written down inside a text terminal of this size.

You would agree, no doubt, that "0.999999999....." is a representation of a real number; and by the ellipsis I am suggesting a pattern for something which is much larger than I would like to write down, although I DO KNOW HOW to write it down -- eg: and that the representation improves in accuracy and may have a selectable accuracy for a given discussion or purpose.

The algorithm itself can be adapted for various constraints; eg: there are many such algorithms based on a similar methodology.

DO you care if I count only the positive reals, or do I need to include the negative as well?
etc.
if you give a recursive enumeration of every real between 0 and 1, that should sufice. such an algorithm, if you indeed possess it and it does what you claim, it would be of great interest to many people.

but...it is highly likely that there is an undiscovered flaw in your algorithm, which might not easily be detectable (especially if the algorithm is of sufficient complexity).

Quote Quote by micromass View Post
A commutative diagram is not a proof. And many mistakes in category theory actually amount to misleading diagrams. A commutative diagram is a suggestion of a proof. It is merely an illustration.

If you look at Bourbaki, then they show that they treat commutative diagrams very rigorously. They try to prove everything with algebraic means.

To me, a proof is correct if it formilizable in some kind of logical system. You don't actually NEED to formalize it, but just giving a suggestion that it can be done is good enough.
commutative diagrams COULD be proofs. for example p→q is just a small diagram. yes, you have a "different kind of formalizing", but logical need not be lingual. the common usage of diagrams now IS informal, this is true, but formal systems can be created which aren't string manipulation. an example can be found in:

http://www.amazon.com/Laws-Form-G-Sp.../dp/0963989901

until relatively recently, geometric proofs ("pictures" a la Euclid), were the gold standard of "formal proofs". it's likely that as we accumulate more knowledge, better formal systems using more sophisticated building blocks will emerge.

(by the way, metamath is an awesome site).

Quote Quote by Hurkyl View Post
Ah joy, "completed infinity".

Just for your information, I have never encountered somebody who uses that terms that could give an even quasi-rigorous account of what the term means.

From observing people who use it, I quite honestly believe that it literally means "I am unwilling or unable to engage in one more layer of abstraction, and thus have failed to recognize that we're talking about the same idea." (I don't mean insult by this, except possibly in the 'unwilling' case)
why, the uniquely defined impossibly inaccessible cardinal, of course :P
andrewr
#72
Feb6-12, 11:41 PM
P: 263
Quote Quote by Deveno View Post
if you give a recursive enumeration of every real between 0 and 1, that should sufice. such an algorithm, if you indeed possess it and it does what you claim, it would be of great interest to many people.

but...it is highly likely that there is an undiscovered flaw in your algorithm, which might not easily be detectable (especially if the algorithm is of sufficient complexity).
It is a simple algorithm. Don't underestimate my abilities. :)

I will tailor it to the needs of the discussion. It is probably easier to do an algorithm that counts all possible reals than one limited to [0-1).

Let me ask a question to make sure we are agreeing on a general idea:
I am going to choose a non-repeating decimal to make sure you don't think I am over-simplifying.

Take the number (2)**0.5. The representation of this real number is given to you by a counting number (2) and a function ()**0.5. I would argue that we know such a number is not finite in representation because there is no FINITE sequence of digits which repeats.
eg: the first digits in a 4 mantissa and 4 fractional representation would be: 0001.4142
By induction, the next symmetrical word size up would be:
00001.41428 ... etc.

It is not a rational number that it should be represented by any ratio of counting numbers.

The proof of the infinite nature of the decimal representation is *inductive* in that no matter how many digits I show you, you can claim there is a representation with n+1 digits.

It is not necessary to write all the digits down to show that it is infinite, but simply to show that it grows without bound. But, none the less, the only way to express this number is a mapping through a function (sqrt) of a counting number. The sequence of digits, however, 1.41428.... is unique.

I am going to assume that a sequence that preserves previously made digits and adds new ones in a predictable (inductive) pattern is sufficient to show the signature of a unique real number; and that likewise, a counting number may be converged to by a signature (in the reverse direction). Such a pattern, that one may work out to any desired number of digits (without END and therfore infinite) is sufficient to show the existence in the decimal system of such a number. Eg: The sequence 1.41428.... no matter what algorithm produces it, so long as it matches the sequence of digits with not a single exception to the algorithm of sqrt(2) ... must be considered to have produced the same rational number.

I let you ponder that for a bit, because I want to give you the best answer I can.
(This is starting to get fun.... I wish people would help me with my simple statistics and propagation of error problems IF I complete this successfully ..... deal? )

And now the question, is the signature of the rational number inductively matched to all possible digits sufficient to indicate that it has been counted? (I am not saying we actually match all infinite digits, but that we show they *WILL* match.)


Register to reply

Related Discussions
Use Cantor's Diagonalization on the set of Natural Numbers? Calculus & Beyond Homework 8
Cantor diagonalization argument Set Theory, Logic, Probability, Statistics 2
Cantor's diagonalization Calculus & Beyond Homework 5
A new point of view on Cantor's diagonalization arguments General Physics 177
A question on Cantor's second diagonalization argument General Math 36