HN Explorer

8 comments

[ 3.1 ms ] story [ 32.0 ms ] thread
naniwaduni
Doesn't this just encode a list of all the primes in a real number, then define a function which extracts each subsequent value? This seems exactly equivalent to defining an array of primes and a function which indexes that array, and passing that off as a function representing all primes. That we can multiplex datagrams into a stream isn't particularly novel.
notfashion
Representing an arbitrary sequence of bits, digits, etc. using a real number is the basis for arithmetic coding. As such, it's an approach that has been productive historically, and arithmetic coding continues to be widely used. So the basic idea of encoding whatever you want to encode in a single real number (obviously represented to just enough precision) isn't fruitless and boring. In this case it is a number that is easy to calculate using the formula given, which doesn't at first glance resemble a prime sieve.
rlpb
> In this case it is a number that is easy to calculate using the formula given

It appears to me that the formula given requires pre-existing knowledge of the list of primes in order the expand the series.

naniwaduni
It is exactly like arithmetic coding. The formula given allows you to improve the precision of the constant given by encoding more primes, and it equally well lets you encode integers that are not prime. It doesn't resemble a prime sieve because it's not one; you just construct approximations with existing knowledge of what the primes are.
AlEinstein
In a sense, this function isn't that interesting; It's just a way of encoding the prime numbers into a real number.

However, one interesting aspect is how efficiently this particular function works for the prime numbers. The function doesn't work for Fibonacci numbers, for example.

Still, it takes 113 bits to encode the primes up to 100, which can trivially be encoded with 100 bits so...