Hacker News Reader

This is in Racket, but it has an arbitrary subjective threshold of 1%: (Online version: http://pasterack.org/pastes/94321 )

  #lang racket
  
  (define (~=~? x y)
    (and (= (length x) (length y))
         (for/and ([vx (in-list x)]
                   [vy (in-list y)])
           (define avg (/ (+ (abs vx) (abs vy)) 2))
           (define diff (abs (- vx vy)))
           (< diff (/ avg 100)))))
  
  (~=~? '(1047072 1047216 1047441 1047521 1047682 59102 59361 59583 59818)
        '(1047085 1047276 1047471 1047754 1047938 59128 59364 59732 59945)) ; ==> #t
  
  (~=~? '(1047072 1047216 1047441 1047521 1047682 59102 59361 59583 59818)
        '(1047085 1047276 1017471 1047754 1047938 59128 59364 59732 59945)) ; ==> #f

Then it's disqualified. 1% is just an arbitrary constant and if it's suitable or not depends on the input. The operator must not have any constants, otherwise it won't work in the general case.

Then how would you define what is "suitable" in the general case?

That depends on the input, and that's what it must be able to tell, unbiased without any pre-defined constants.

I've thought a bit more and made a small but important change in how the "equalish" operator works:

The operator should take three (not two as suggested before) sets of integers as input of equal sample size.

The first two sets are already known to be equalish TRUE, as defined by either a human or a previous test of a smaller subset of the integer sets. The third set is a the new sample you want to compare.

Example output:

equalish('{0}','{0}','{0}') -> TRUE

equalish('{0}','{0}','{1}') -> FALSE

equalish('{110,105,120}','{113,107,121}','{110,106,120}') -> TRUE

equalish('{1047072,1047216,1047441,1047521,1047682,59102,59361,59583,59818}','{1047085,1047276,1047471,1047754,1047938,59128,59364,59732,59945}','{1047085,1047276,1017471,1047754,1047938,59128,59364,59732,59945}') -> TRUE

Which of first or second input are we comparing the third to? If both, then I don't think you operator makes sense, since that would suggest your ~=~ operator is transitive. Since all inputs are comparable, I assume, that would mean your operator is an equivalence relation.

Due to transitivity:

A ~=~ B ~=~ C and

C ~=~ E ~=~ F implies that

A ~=~ F which would not necessarily be the case.

For instance:

Let X be a large set with many elements. Then a "reasonable intelligent human" might suggest that:

(X `union` {1}) ~=~ (X `union` {2})

And also:

(X `union` {2}) ~=~ (X `union` {3})

...... (X `union` {999999999999}) ~=~ (X `union` {1000000000000 })

which implies

(X `union` {1}) ~=~ (X `union` {1000000000000 })

, even if this is actually the case, you expect it not to be for some large enough number. Then your operator cannot be transitive, but then what use is the knowledge that input #1 is approximate equal to input #2?

But since you already "discovered" the algorithm, why not just show it, or the pseudo code, such that we could discover its true name?

You can create an operator that can be made of other operators

cardinal of A intersection B == cardinal of (A) (or cardinal of B : since A~=~B implies card( A )= card( B) ) This is your test re-expressed by using math.

But basically the formula I gave is the shortest non factorable form of this operator in a normal set theory.

And there you have a problem, every truth are biased by a corpus of prejudice called theory and the physical nature : because of the nature of computer you cannot represent any given number. There always be a number that can saturate your computer resources. It is called MAXINT.

Old MAXINT where the size of a register. With Big Int there are the size of your allocatable computer memory (phys + virt)..... on a distributed grid it will be the resulting union of resources. And universe being resource bound there always be a limit to the representation of the integer you can make. So computers will still fall short of resources to represent ALL numbers.

So the absolute operator you call for does not and cannot exists as long as computer are resource bound. So there is a map problem here that cannot draw completly the territory. The playground of physic is bounded, math abstraction/measures are not.

And your way of interpretating data can be wrong : what tells you the input number are not modulo something? Like in a % 2 universe : 2 == 0 [2].

Good software are high context software. There is no such thing as immutable truth and monads is still an elucubration of a monk that ended being burnt by the church for blasphemy. Not to say he was wrong on saying the earth is moving (in fact the solar system)... but to say ignoring context can literally burn you.

PS Noise is defined by the observator. It is measuring the log of the ratio of non of relevant choices over relevant choices (entropy) and entropy varies according to what the observer calls relevance of the data.

What is relevant to someone (random numbers in cryptography) maybe irrelevant to other ones.

Define 'noisy' and then what you mean by two sets being 'equal', given that definition.

I don't get it. The last two examples in your post compare the exact same sets, yet the result of the operator is TRUE in the first example, and FALSE in the second? Or is this a joke that's gone over my head?

the second sequence in the first example starts: 1047085,1047276,1047471

in the second example it starts: 1047085,1047276,1017471

The 3rd element is different.

Why is the last example false? I'm not sure what you mean by equal in this context, I guess.

Common Lisp: equalp

No. That's not the one. "Two objects are equalp if they are equal; if they are characters and satisfy char-equal, which ignores alphabetic case and certain other attributes of characters; if they are numbers and have the same numerical value, even if they are of different types; or if they have components that are all equalp."

Introducing the ~=~ operator