Is there a strictly algebraic way of coding "whole number value of"?

In summary, the conversation discusses the possibility of coding the notion of "whole number value" for an irrational number using only algebraic symbols or numbers. The idea of using a recursive algorithm to check for the integer part of a square root is proposed, but it is acknowledged that this may not be efficient for all numbers. It is also noted that it is not possible to have an algebraic closed formula that gives the whole number part of most numbers, as the algebraic numbers make up only a small portion of the real numbers.
  • #1
David Carroll
181
13
Greetings in the name of Jah-gonaut, ladies and gents. I have a question. Is there a strictly algebraic way of coding the notion "whole number value of", specifically for an irrational number?

I know that one way of coding the notion "absolute value of" some number is taking the square root of the square. Can something similar be done to code for "whole part" of some number, using only algebraic symbols or numbers? Namely, in my case, can some kind of manipulation of symbols be done with the irrational part of a solution to some polynomial equation {by which I mean the "sqrt(b2 + 4ab)" part of the solution)}, using ONLY the symbols a or b?

For example, let's take the two solutions to the polynomial equation x2 + 21x - 155 = 0:

-26.7864974749... and 5.7864974749...

I want to separate the decimal expansion from the whole value 5 for the second solution. But I want to perform this step using only algebraic symbols, pretending - say - that I'm some machine that doesn't otherwise know the command "separate the whole part from the decimal expansion".

Can I perform some finite manipulation of the symbols a, b, or c (from the polynomial formula) to do this?
 
Mathematics news on Phys.org
  • #2
Correction: "sqrt(b2 + 4ab)" should read "sqrt(b2 + 4ac)"
 
  • #3
Actually, correction again: that is "- 4ac", not "+ 4ab". And that is, quadratic formula, not polynomial formula.
 
  • #4
Assume we're only dealing with real numbers.

Let ##\Phi_0(z)\equiv##"##\exists a,b\in\mathbb R \text{ such that } z = a^2, \enspace 1-z = b^2 \text{ and } b\neq 0##." It's straightforward to check that ##\Phi_0(z)## is true if and only if ##z\in[0,1)##.

For any ##n\in \mathbb N##, let ##\Phi_n(z)\equiv \Phi_{n-1}(z-1)## and ##\Phi_{-n}(z)\equiv \Phi_{-(n-1)}(z+1)##.

It's straightforward to check that ##\Phi_n(z)## is true if and only if ##\lfloor z\rfloor=n##.

I would argue that the Boolean ##\Phi_0## is just as algebraic as ##\sqrt{\enspace}## is. Given that, you could get a computer to (recursively) check ##\Phi_n, \Phi_{-n}## for each ##n##, and it will eventually terminate and find your integer part.
 
  • #5
How efficient is such an algorithm on a computer? Quicker than, say, shor's algorithm?
 
  • #6
I guess, since it's recursive, it would be just as efficient as taking the square root of a real number?
 
  • #7
Every recursive algorithm can be transformed to an iterative algorithm. The recursive algorithm that has been proposed amounts to iteratively checking whether ##i^2 > x >= 0## for each non-negative integer i in order until one is found that satisfies the test. The floor of the square root of x is given by the first i that passes the test.

That algorithm takes approximately 10^5 steps to determine the floor of the square root of 10^10 - 1.

It is possible to compute the square root of 10^10 - 1 to 5 significant digits in WAY fewer steps than that.
 
  • #8
Suppose you can apply ##+,-\times,\div##, and roots to some irrational ##\alpha## (and algebraic numbers) to get a whole number ##n##. Then you get ##R(x)-n## is an algebraic function with ##\alpha## as a zero (where ##R(x)## was the closed formula used). Consequently, ##\alpha## is algebraic.
So it is not possible to have an algebraic closed formula (as opposed to recursive algorithm) that gives the whole number part of most numbers (since the algebraic numbers are barely any of the real numbers).
 
  • #9
This is disappointing news. But thank you.
 

What is the definition of "strictly algebraic way of coding"?

A strictly algebraic way of coding refers to a method of representing numbers or data using only algebraic operations, such as addition, subtraction, multiplication, and division, without the use of any other symbols or functions.

Can whole numbers be represented using only algebraic operations?

Yes, whole numbers can be represented using only algebraic operations. This is because whole numbers are a subset of the set of real numbers, and algebraic operations can be used to manipulate and represent all real numbers.

What are some examples of algebraic coding for whole numbers?

One example of algebraic coding for whole numbers is the use of variables and equations to represent numbers. For instance, the equation 2x + 3 = 7 can be used to represent the whole number 2.

Can algebraic coding be used for other types of numbers besides whole numbers?

Yes, algebraic coding can be used for other types of numbers, such as fractions, decimals, and even irrational numbers. The key is to use algebraic operations to manipulate and represent these numbers in a meaningful way.

What are the advantages of using a strictly algebraic way of coding?

Some advantages of using a strictly algebraic way of coding include its simplicity and consistency, as well as its ability to represent a wide range of numbers and data. Additionally, algebraic coding can be easily translated into computer code, making it useful for programming and data analysis.

Similar threads

Replies
7
Views
1K
  • Engineering and Comp Sci Homework Help
Replies
21
Views
2K
Replies
2
Views
2K
Replies
8
Views
1K
  • Calculus and Beyond Homework Help
Replies
32
Views
2K
Replies
1
Views
2K
Replies
1
Views
3K
Replies
1
Views
2K
  • Programming and Computer Science
Replies
1
Views
952
  • Linear and Abstract Algebra
Replies
1
Views
988
Back
Top