On Sun, Dec 29, 2013 at 4:51 PM, meekerdb <meeke...@verizon.net> wrote:

>  On 12/29/2013 1:28 PM, Jason Resch wrote:
> On Sun, Dec 29, 2013 at 2:25 PM, meekerdb <meeke...@verizon.net> wrote:
>>  On 12/29/2013 5:56 AM, Bruno Marchal wrote:
>>  On 28 Dec 2013, at 22:23, meekerdb wrote:
>>  On 12/28/2013 4:09 AM, Bruno Marchal wrote:
>> For a long time I got opponent saying that we cannot generate
>> computationally a random number, and that is right, if we want generate
>> only that numbers. but a simple counting algorithm generating all numbers,
>> 0, 1, 2, .... 6999500235148668, ... generates all random finite
>> incompressible strings,
>> How can a finite string be incompressible?  6999500235148668 in base
>> 6999500235148669 is just 10.
>>  You can define a finite string as incompressible when the shorter
>> combinators to generate it is as lengthy as the string itself.
>> This definition is not universal for a finite amount of short sequences
>> which indeed will depend of the language used (here combinators).
>>  Then you can show that such a definition can be made universal by
>> adding some constant, which will depend of the universal language.
>>  It can be shown that most (finite!) numbers, written in any base, are
>> random in that sense.
>>  Of course, 10 is a sort of compression of any string X in some base,
>> but if you allow change of base, you will need to send the base with the
>> number in the message. If you fix the base, then indeed 10 will be a
>> compression of that particular number base, for that language, and it is
>> part of incompressibility theory that no definition exist working for all
>> (small) numbers.
>>  Since all finite numbers are small, I think this means the theory only
>> holds in the limit.
>> Brent
>  Brent,
>  It is easy to see with the pigeon hole principal.  There are more 2
> digit numbers than 1 digit numbers, and more 3 digit numbers than 2 digit
> numbers, and so on.  For any string you can represent using a shorter
> string, another "shorter string" must necessarily be displaced.  You can't
> keep replacing things with shorter strings because there aren't enough of
> them, so as a side-effect, every compression strategy must represent some
> strings by larger ones.  In fact, the average size of all possible
> compressed messages (with some upper-bound length n) can never be smaller
> than the average size of all uncompressed messages.
>  The only reason compression algorithms are useful is because they are
> tailored to represent some class of messages with shorter strings, while
> making (the vast majority of) other messages slightly larger.
> A good explanation.


> But just because you cannot compress all numbers of a given size doesn't
> imply that any particular number is incompressible.

That is true if you consider the size of the compression program to be of
no relevance.  In such a case, you can of course have a number of very
small strings map directly to very large ones.

>   So isn't it the case that every finite number string is compressible in
> some algorithm?  So there's no sense to saying 6999500235148668 is random,
> but 11111111111111 is not, except relative to some given compression
> algorithm.

Right, but this leads to the concept of Kolmogorov complexity. If you
consider the size of the minimum string and algorithm together, necessary
to represent some number, you will find there are some patterns of data
that are more compressible than others.  In your previous example with base
6999500235148668, you would need to include both that base, and the string
"10" in order to encode 6999500235148669.  For the majority of numbers, you
will find the Kolmogorov complexity of the number to almost always be on
the order of the number of digits in that number.  The exceptions like
1111111111 are few and far between.


You received this message because you are subscribed to the Google Groups 
"Everything List" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to everything-list+unsubscr...@googlegroups.com.
To post to this group, send email to everything-list@googlegroups.com.
Visit this group at http://groups.google.com/group/everything-list.
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to