Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
Let's say you symbols are a number of dots. The distance between successive dots is used to carry information. If for a minute, we ignore that we are in a quantum reality, and that you can place the dots anywhere along a straight line, and we can measure the distance without error, then we can use the distance to represent any real number (as you can have any real distance between the dots). Note: in this scenario, even if we limited the maximum distance between successive dots to be, say 10 cm, we could still represent all reals with only two dots. More dots are not strictly necessary.
Measure once. Cut twice.
Player (36)
Joined: 9/11/2004
Posts: 2630
Warp wrote:
Scepheo wrote:
If your set of symbols is uncountable you can represent the reals just fine.
Could you explain how a set of distinct individual symbols (that can be used to represent numerical values) can be uncountable? Because I cannot even fathom how. (Granted, I'm not that good at math and grasping things like infinite sets, but still...)
You can count an infinite set if you can map it to the whole numbers. By this logic it's possible to say that you can count all of the integers because you can map them as follows.
0, 1,  2, 3,  4, 5,  6, 7...
0, 1, -1, 2, -2, 3, -3, 4...
You can also count all rational numbers by mapping them as follows:
1/1 1/2 1/3 1/4 1/5 1/6...
2/1 2/2 2/3 2/4 2/5 2/6...
3/1 3/2 3/3 3/4 3/5 3/6...
4/1 4/2 4/3 4/4 4/5 4/6...
5/1 5/2 5/3 5/4 5/5 5/6...
6/1 6/2 6/3 6/4 6/5 6/6...
Then start from the top corner and perform a trick called diagonalization. You visit each node by starting at 1/1, then visiting 2/1 then 1/2, then 3/1, 2/2 (skipped, members not coprime), 1/3, then 4/1, 3/2, 2/3, 1/4, etc. Doing this you can visit each possible rational number in a methodological manner. But this doesn't work for irrational numbers. Let's say that I have a list of real numbers (between 0 and 1) that I say maps to whole numbers, and it looks like this:
0: 0.487397268962....
1: 0.798567386647....
2: 0.982764976473....
3: 0.496739863834....
4: 0.875976387663....
5: 0.095479836476....
6: 0.798463876374....
7: 0.129828732002....
I can make a new number, that's not on this list, by doing the following: Take the 1st digit from the first number, change it, take the second digit from the second number, change it, and continue, by collecting a different digit from each number. Now you have a number (one possibility is: 0.5038809487239...) that is a real number, and is not on the list. Therefore, your complete list, is actually not a complete list. Therefore, you cannot count all real numbers. Any method you try will miss some. In fact, any method you try will miss infinitely many, and the amount that it will miss is also uncountably many.
Build a man a fire, warm him for a day, Set a man on fire, warm him for the rest of his life.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
OmnipotentEntity wrote:
Therefore, you cannot count all real numbers.
I know Cantor's diagonal argument, and your post answered the question "can you prove that there are uncountable sets?" But that was not the question I asked.
Editor, Expert player (2073)
Joined: 6/15/2005
Posts: 3282
Warp wrote:
Could you explain how a set of distinct individual symbols (that can be used to represent numerical values) can be uncountable?
Declare each real number to be its own symbol, distinct from the symbol of any other real number. Clearly, the symbols as a set are distinct individual symbols, representing numerical values (through a bijection with the reals) and uncountable (since they are in bijection with the reals). If "Declare each real number to be its own symbol" is still confusing, think about giving each real number r a symbol A_r. For example, the real number 1 has symbol A_1, the real number π has symbol A_π, and so on (π is pi, by the way; it shows up in a rigid and unfamiliar form on my screen).
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
FractalFusion wrote:
Declare each real number to be its own symbol, distinct from the symbol of any other real number.
Real numbers cannot be symbols because they cannot be all represented. What you did is just a tautology: "I represent each real number with itself." That leads us nowhere. Of course the set of real numbers is uncountable. I already said that.
Joined: 5/2/2006
Posts: 1020
Location: Boulder, CO
What do you mean by symbols then? given r a real number, we can write a symbol for it. There are uncountably many symbols in this naming scheme, so it should not surprise that each real has a symbol in the scheme.
Has never colored a dinosaur.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
Twelvepack wrote:
What do you mean by symbols then?
The whole concept of "represent a number" is vague. I suppose I'm talking about this.
given r a real number, we can write a symbol for it.
Can you actually prove that? (Doesn't it assume the veracity of the axiom of choice?)
Player (36)
Joined: 9/11/2004
Posts: 2630
What constraints do you impose upon mathematical symbolism? That it must be expressible in this universe? If that's the case then we cannot even construct symbols for each real number. Because to do that would require infinite space. For instance, how would you represent a not-special number in the neighborhood of Graham's number? If we eschew the requirement that it be representable in the universe then we can simply represent each number by a line segment with a length in any sufficiently well defined unit system.
Build a man a fire, warm him for a day, Set a man on fire, warm him for the rest of his life.
Player (80)
Joined: 8/5/2007
Posts: 865
I just took a minute to browse discussion of the latest question and I have a related question (I think). Is there a system for representing any number uniquely and unambiguously that does not include exceptions? There are two systems that I'm familiar with, both of which fail to offer unique representations: • The traditional decimal system (or any other base, for that matter). In this system, any terminating decimal can be represented as the same number minus one in the last digit, followed by trailing nines. For example, 0.123 can be just as easily represented as 0.122999999... Therefore, the decimal system doesn't uniquely define the number 0.123. • The continued fraction system. In this system, any rational number can be represented in two ways: as either {..., an} or {..., an-1, 1}. For example, 5/64 can be equivalently represented as either 1/(12+1/(1+1/4)) or 1/(12+1/(1+1/(3+1/1))) (in Wikipedia's notation, these two representations would be denoted [12, 1, 4] and [12, 1, 3, 1] respectively). Mathematicians here may leap in and say, "The solution is simple! Just disallow the representation that isn't preferred!" That's why I've added the condition that there are to be no exceptions. After all, we know what 0.122999999... is telling us just as well as we know what 0.122444444... is telling us; each is just an infinite sum of coefficients of powers of ten. The same is true of ambiguities in the continued fraction notation-- just because you disallow [12, 1, 3, 1] doesn't mean you don't know how to interpret it. Can this be done?
Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
Using any language composed of a finite number of countable symbols, we can't express any real number anyway, there are just more numbers than there are combinations of symbols, however, we can get as close as we want to any number. If we are only concerned about getting as close as we need to, then the decimal system should work just fine. 0.9999999999... is not a valid representation of a number in a strictly defined decimal system. Why? Because in order to represent that the number is recurring, we needed to add an ellipsis. Strictly speaking, we can say the ellipsis is not a defined symbol, therefore, 0.999999999... is not a valid representation of the number 1, and 0.999999999 is a different number than 1. We can get as close to any number you want with only a finite number of symbols. If we have an uncountable number of symbols, then you could just give each real number it's own unique symbol, therefore satisfying your criteria.
Measure once. Cut twice.
Player (80)
Joined: 8/5/2007
Posts: 865
So... I've returned the argument to where it was about a dozen posts ago? I feel like there's a nontrivial way to ask this question. How about a system that represents only the rationals uniquely with a finite number of symbols? There seems to be enough trouble with those alone. Would it also help if I demand the numbers be defined algorithmically? (I'm not sure how to define "algorithmically", but I propose it just to get around your suggestion that we define each number as a collection of symbols. Maybe "algorithmically" can be defined such that the symbols can be truncated at any point to produce an approximation which gets more precise with each successive symbol.) Also, I take issue with your discarding 0.99999... because of the ellipsis. Trailing zeroes are implied at the end of any terminating decimal-- otherwise, we could represent 0.123 as 0.1230, 0.12300, 0.123000, and so on. Thus, I demand that every number be represented as an infinite sequence of digits. This forces a choice strictly between 0.1230000... and 0.12299999...
Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
Here's a system which I believe fits your criteria (or at least gets close) Each number can be represented by a sign plus a set of integers. How it works? The nth integer in the series represents the power in which you take the nth prime to get your number. Decimal points and numbers afterwards are not defined in this system. We assume infinite trailing zeroes. let's define 1 as negative, 0 as positive and 2 as zero. A number such as 5 would be represented by [0, 0, 0, 1] = ( 1 * 2^0 * 3^0 * 5^1). A number like -12 would be [1, 2, 1] = ( -1 * 2^2 * 3^1 ) A number like 2.5 would be [0, -1, 0, 1] = (1 * 2^-1 * 3^0 * 5^1) There is a unique representation for every rational number. Taking all the primes with negative exponents, we get the denominator of the fraction. Taking all of the primes with positive exponents, we get the numerator. They must be coprime because they have no prime factors in common. We also know that there is only one unique representation for any rational number using a fraction where both the numerator and denominator are coprime. Unfortunately this method does not have a very good way to express 0. Introducing a third number for the sign can fix this However, that means you can put whatever you like for anything past the first 2, however if you do decide to remove the sign digit, it does provide a way to express all positive rational numbers uniquely. Maybe this isn't exactly what you're looking for, but I think it gets pretty close. EDIT: Actually, there is a simpler way which also includes 0. Define all rational numbers in this way: One integer and one natural number. The integer will represent the numerator of the fraction. It can be any number, even negative, but it must be an integer. The natural number n is the nth smallest natural number that is coprime with the integer described above. For example, 5 would be [5,1] = (5 / 1 (1 is the first coprime number to 5)) , -3.33333 would be [-10, 2] = (-10 / 3 (3 is the 2nd natural number coprime to 10)). 0 is [0,1]. There is no other way to express 0 because there is only one natural number coprime to 0, an that's 1. [0,2] is not valid as it implies there is a second natural number coprime to 0.
Measure once. Cut twice.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
OmnipotentEntity wrote:
If we eschew the requirement that it be representable in the universe then we can simply represent each number by a line segment with a length in any sufficiently well defined unit system.
I hope you understand why that sounds so unsatisfactory as an answer, even though it's difficult to define "representation" in this context. A small amount of symbols (in fact, two) is enough to represent all integers and even rational numbers, and the representation is trivial to define and to interpret. It does not require supernatural accuracy. (Even though such a representation is arbitrarily large, any given integer or rational number has a finite representation that could in principle be trivially interpreted, simply by reading the sequence of two symbols.) Your suggestion is extremely non-trivial, requires an infinite amount of symbols and, basically, requires infinite accuracy in order to be interpreted correctly (which is quite unlike the representation of integers, which is always finite and trivial to interpret, no matter which integer you choose.) It's also dependent on the physical shape of the symbol (something that integers/rationals do not require at all.) Can it be called a "symbol" at all?
Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
When we say symbol, we are talking about individual digits of a number, a decimal point, or an addition sign for example, and not numbers themselves. If you are talking about referring to rational numbers of the form a/b, and saying that requires only two symbols, you are mistaken (unless you have an infinite alphabet). It only requires one integer to represent all rational numbers anyway, since the number of rational numbers is countable.
Measure once. Cut twice.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
andymac wrote:
If you are talking about referring to rational numbers of the form a/b, and saying that requires only two symbols, you are mistaken
Ok, it requires three: Two digits and a way to separate the numerator from the denominator. (Although you can use a sequence of digits to first tell how many digits are in the numerator, and end it with a unique sequence. It's simpler if you just use three symbols.)
(unless you have an infinite alphabet).
Where's that coming from?
It only requires one integer to represent all rational numbers anyway, since the number of rational numbers is countable.
But you still need (at least) two digits (ie. symbols) to represent the integer.
Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
So far there are only 10 numbers, 26 letters of upper and lower case and however many other special symbols you want to add to your alphabet. If you want to represent all the integers with only three symbols, you're going to have a bad time. You would need the aleph-naught members in your alphabet in order to represent all the integers if you only wanted to use 3 letters/numbers/characters. Or are you talking about how many unique symbols you need? In which case, you can use one symbol to represent all rational numbers.
Measure once. Cut twice.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
andymac wrote:
So far there are only 10 numbers, 26 letters of upper and lower case and however many other special symbols you want to add to your alphabet. If you want to represent all the integers with only three symbols, you're going to have a bad time. You would need the aleph-naught members in your alphabet in order to represent all the integers if you only wanted to use 3 letters/numbers/characters. Or are you talking about how many unique symbols you need? In which case, you can use one symbol to represent all rational numbers.
I don't even understand what you are talking about there. There's some kind of miscommunication issue at play here. "If you want to represent all the integers with only three symbols, you're going to have a bad time." "you can use one symbol to represent all rational numbers." This seems to be saying that you can't represent all integers with three symbols, but you can represent all rational numbers with one symbol. Pardon me if I find this highly contradictory.
Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
Let me disambiguate this situation. I use "Symbol" to mean two different things. So I'll introduce two new words, let's use "digit" and "character". Digit will represent length, and character will represent type. For example, the number 1111 is four digits long, but it only uses one character. We can represent all rational numbers using one character. However, we cannot place a limit on the number of digits. If you tried to represent all of your integers using only three digits, you would need an infinite selection of characters (which is what I meant by an infinite alphabet). If you place a limit on the number of characters, then you will need a potentially infinite number of digits to represent all of your numbers. However, if your number is finite, then the number of digits required to represent your number will be finite as well.
Measure once. Cut twice.
Banned User
Joined: 3/10/2004
Posts: 7698
Location: Finland
andymac wrote:
If you tried to represent all of your integers using only three digits, you would need an infinite selection of characters
I didn't say that you can represent all integers with three digits. I said you can represent all integers (and rational numbers) with three symbols.
If you place a limit on the number of characters, then you will need a potentially infinite number of digits to represent all of your numbers.
So what? That's not the point. The point is that any integer (or rational number) can be represented with a finite sequence of symbols (from a finite set of them). There's no integer that would require an infinite amount of them. (Also, this representation is not dependent on ancillary things like the size or shape of those symbols.) The same cannot be said of real numbers. You cannot represent every single real number with a finite sequence of symbols (from a finite set).
Experienced player (623)
Joined: 11/30/2008
Posts: 650
Location: a little city in the middle of nowhere
Okay, so we agree... what's your point?
Measure once. Cut twice.
Editor, Expert player (2073)
Joined: 6/15/2005
Posts: 3282
A couple problems involving Catalan numbers. Let Cn denote the nth Catalan number. 1) Find all k≥0 for which Ck divides Ck+1. 2) Prove that Ck is odd if and only if k=2j-1 for some integer j≥0.
Player (42)
Joined: 12/27/2008
Posts: 873
Location: Germany
1) C_(k+1) = (2k+2)!/((k+2)!(k+1)!) = (2k+2)(2k+1)/(k+2)(k+1) * C_k So, C_k divides C_(k+1) if that fraction is an integer. Simplifying it we have 2(2k+1) = m(k+2) => (4-m)k = 2(m-1) => k = 2(m-1)/(4-m) Since k>=0, m<4 and m>=1 and we can simply test m=1,2,3. k is integer for all these 3 and we have k=0, k=1 and k=4. So, C_0 divides C_1, C_1 divides C_2 and C_4 divides C_5, no more cases like this occur. 2) C_k = (2k)!/k!(k+1)! = 1/k+1 (2k k) = (1 - k/(k+1)) (2k k) = (2k k) - (2k k+1) The (n m) denotes the binomial "n choose m". We have to calculate C_k mod 2, so just apply Lucas' theorem to the binomials. If we express k using binary digits, 2k is k shifted to the right with a 0 appended to the end. From Lucas' theorem, (2k k) is even if, for some position d, the d-th digit on 2k is 0 and the d-th digit on k is 1, because (0 1) = 0 and all the others are 1. Now, assume k>0 and pick the least significant bit of k that's 1, the bit on the same position on 2k will be either 0, if k is the rightmost bit, or the bit to the right, which by our choice should be 0 too. This proves that (2k k) is even for k>0. Now for (2k k+1). Also assume k>0 and pick the least significant bit of k that's 0, at position d. When you calculate k+1, this bit becomes 1, all others to the right become 0 and all others to the left don't change. From our choice, the bit at position d on 2k is 1, and we get (1 1) = 1. To the right, k+1 has only 0 bits, so we get (x 0) = 1, so no problems there. Now pick, if it exists, the rightmost bit of k that's 1 and is to the left of d, at position e. The bit on position e at 2k is necessarily 0, and since adding 1 to k doesn't change the bit at position e, we get (0 1) = 0 and (2k k+1) is even. If this bit doesn't exist then k is represented by a string of 1's, so for (2k k+1) we would have (1 1) in the leftmost bit, (0 0) in the rightmost and (1 0) for all others, and (2k k+1) is odd. If k is represented by a string of 1's, then k = 2^j - 1, j>0. For this, we have (2k k) even and (2k k+1) odd, so the subtraction, C_k is odd. For all other k>0, both (2k k) and (2k k+1) are even, so C_k is even. It's trivial to check that C_0 is odd. Therefore, we conclude C_k is odd iff k = 2^j - 1, j>=0 EDIT: I solved the "if" part of 2) without using Lucas' theorem: By counting the number of multiples of 2, 4, 8, 16, ... we get the number of factors of 2 in n!: sum(floor(n/2^k),k), k>=1 We use this to find how many factors of two are in (2n n) = (2n)!/(n!)² So, there are sum(floor(2n/2^k),k) - 2*sum(floor(n/2^k),k) = sum(floor(2n/2^k)-2*floor(n/2^k),k) factors of two, if we let n/2^k = floor(n/2^k) + a_k and 2n/2^k = floor(2n/2^k) + b_k, we have: floor(2n/2^k) - 2*floor(n/2^k) = 2a_k - b_k a_k = (n mod 2^k)/2^k b_k = (2n mod 2^k)/2^k We have b_1 = 0 and b_k = a_(k-1). The k-th term of the sum is (n mod 2^k - n mod 2^(k-1))/2^(k-1). If n = 2^j - 1 then n mod 2^k = 2^k - 1, if k<=j, for k in that range the term reduces to (2^k - 2^(k-1))/2^(k-1) = 1, there are exactly j terms in that range. So the sum evaluates to j, meaning that, when n=2^j-1 there are exactly j factors of 2 in (2n n). C_n = (2n n)/(n+1) = (2n n)/2^j, that means we take off all j factors of two, and an odd integer is left over. I think I can prove the converse if I assume n + 1 = m*2^e, with m an odd integer, and prove that the number of factors of two in (2n n) is larger than e, maybe I'll do it later.
Editor, Expert player (2073)
Joined: 6/15/2005
Posts: 3282
For 2, I would just use induction and the recursion property: ● C0=1, Cn+1=Sum(i=0 to n)CiCn-i We can write (for k>0): ● C2k=2*Sum(i=0 to k-1)CiC2k-1-i ● C2k-1=Ck-12+2*Sum(i=0 to k-2)CiC2k-2-i So C2k is always even and C2k-1 has the same parity as Ck-1. The statement is true for n=0 since C0=1 and 20-1=0. Suppose the statement holds for all n<m where m>0. If m is even, then Cm is even by above, and m cannot be written as 2j-1. Otherwise, m is odd. Let k be such that m=2k-1. Then the following statements are equivalent: ● Cm=C2k-1 is odd ● Ck-1 is odd. ● k-1=2j-1 ● m=2k-1=2(k-1)+1=2(2j-1)+1=2j+1-1 so Cm is odd if and only if m is of the form 2i-1. Therefore the statement holds for n=m. This completes the induction.
Player (42)
Joined: 12/27/2008
Posts: 873
Location: Germany
Nice induction! I managed to finish my solution with pure number theory. I found that the amount of factors of 2 in (2n n) is the sum of (n mod 2^k - n mod 2^(k-1))/2^(k-1). The converse is simple, the terms of the sum are just the digits of n expressed in base 2! In fact, I've proven p4wn3r's theoremTM: Let s be the sum of the digits of n in base 2, then 2s || (2n n) The symbol || means that 2s is the largest prime power of 2 that divides (2n n). With this result, it's simple. We analyze the parity of Cn = (2n n)/n+1. Pick, if it exists, the rightmost bit that's 0. When you evaluate n+1, that bit becomes 1 and all to the right become 0. Say the number of bits to the right is d, so 2^d || n+1. We know 2^s || (2n n) and s is the sum of bits of n, but since the d rightmost bits of n are 1, s>=d. Assuming the bit we picked exists, either n=0 (C_0 odd) or there's at least one more bit that's one, the most significant bit of n, so s>d, and the numerator has more factors of two than the denominator, and C_n is even. If the bit doesn't exist, then n is 11...1 = 2^j - 1 for some j. In this case s=d, all the factors of 2 cancel out and C_n is odd. Similar proofs should also answer which Catalan numbers are not divisible by 4, 8, etc. The ones not divisible by 4 have the form 2^a + 2^b - 1, a>b. By 8, 2^a + 2^b + 2^c - 1, a>b>c and so on.
arflech
He/Him
Joined: 5/3/2008
Posts: 1120
If you look at my avatar, you'll notice that from each vertex of each square, a line segment is drawn in its interior at a certain angle (0 to 45 degrees) clockwise from one of the sides; the points of intersection of pairs of these line segments are the vertices of the next square in the series. As this angle (t) varies, how do the positions of the vertices of the interior squares vary? I first thought this through for the first square with Cartesian coordinates, setting the outermost square to have one vertex at the origin and the opposite vertex at (1,1), and it was elementary, but I suspect it would be easier to use polar coordinates and center the outermost square at the origin. The older version of my avatar is below, for reference:
i imgur com/QiCaaH8 png