Re: Was: Re: What is the state-of-the-art analysing hardware impact on achievable compression ratio


Claudio Grondi wrote:
Nice to see a continuation of the thread I have started on 2005-11-12. I

Wish I could enforce threads remaining threaded. Not so.
Loss of continuity is a major problem with th web.

Which ... continues ... to be a problem. Pub intended.

am very glad, that someone detected the need of further discussion as I
mean the subject and its various aspects are worth to be coped with here.

The topic is non-trivial: and solutions to it are fundamentlly seminal
in terms of fundamentals of IS/IT versus cognitie semiotics.


As I had a very hard time to get through the text of the message below I

Sorry for typos and bad formatting. Apologies that my style of
presettion
may be difficult to follow. I wish that threaded responses could be
more
esily constructd; being limited to only text with indentation to convey
a complex ideational structure (my _thoughts_ and _emotions_)
I also fail like others to convey the ideas precisely or well.

Since you speak more than one language, your educational level
is commensurtly higher than mine, as well as making it more
difficult
for you to understnad and me to communicate: a compliment in
your directon and a tacit warning to me and anyone trying to
communicate
on EITHER end: when it works it is fantastic; but when it fails it
is not
necessarily disasterous or a big negative event: just try again.

hope, that eventual responses put some more structure into it (e.g. in
form of a summary) making it easier to see the main points.

Yes: even as the main points may perhaps become redefined by the
comversation.


Here my attempt towards above:

Be patient with us, as well as yourself: what you see may be the
only
expression of it from the only person seeing it: that beng you.

Feel free to directly correspond (roy.crabtree@xxxxxxxxx) as well as
in
the news group.



The subject of this posting, as I perceive it, is to show some more
aspects worth to be discussed and to animate eventual responders to cope
for this reason with following questions:

One of many. If you & I come to a rsult which improves compression;
or any other constructive beneficent result, also acceptable.



- Where comes the "redundancy" of data making compression possible from?

ALL data is compressible. ALL data is also INcompressible.

PROBABLY it IS the case that

NOT all data is BOTH compressible and INcompressible

using the same dictionary or in exactly the same aspect or
context.

Once youy have data ( a representation) you have a name for it (the
actual data,
in case of classic "scientific" data).

Once you have a name, you can have ANOTHER name (alias).

This can be made as long, or as short (abbreviation) as you wish.

This is compression. (In one sense, not necessarily the only one).

It may be possible on ONE contet to use abbeviations so as to reduce
the overalll "size" (measurement or metric of eigenstate) relative to a
PRESUMED parent context (the "dictionary" of the "language" or
_presumed_
system eigenstate already present).

It may or may NOT be possible to reduce all such conversations;
current proof processes assert it is not.

I tend to agree; but am willing to listen to a proof processes
attempting
to denote tht this property ma also be false.

New theories tend to override older theories that the new theories
"prove" as false.
The _correct_ ones tend to be succesful. (NOT the other way around:
Being successful is NOT necessarily a proof of being CORRECT).

As such, fater than light trel may be found (tachyons);
Godel's Theorem and Cantor Transfinistesimals Theora are correct

but are usually intepreted incorrectly:

Math CAN be consistent and complete
There ARE unnamable numbers BUT the diagonalization argument
does NOT prove OR proof it well.

- When makes it sense to try to reduce the "redundancy" of data
and when not?

First, define redundancy.

Of curse (not a typo), you first have to define _data_.

Which in, and of, itself, is VERY difficult to do WITHOUT repetition:

YOu have no _meaning_ in the _data_ ou have UNLESS
you see the same thing _occur_ again.

Or else you can assign no _meaning_ to it.

YOu have a problem labelled as a primitive recursive kernel:

All the parts have to relate and be defined relative to each other
all at the same tme, even to get started.

If you have to have repetition (also known as _redundancy_)
in order to even get your first datum, ... how do you start?


- What is the right, proper and most useful context of describing data
as compressible or incompressible?

Data will be described as compressible or incompresible depending on
which
context (or dictionary) you adopt for the prpose of attempting to
compress it.

The trick is: in the context tht you )wh_ to compress data, which
additional
data (or contet) doo you need, to be able o do it as effectivey as it
CAN be done
within the contet ou are fored to start with _and_ maintain while doing
it?

-- Communication of textual messages over the network?
-- Storage of measurements results gained from observations?
-- Passing of gained experience, wisdom and achieved technology level
for usage by next generation of humans?
-- Optimizing energy required for information exchange between
specific electronic systems needing coordination?
-- Minimization of efforts towards prediction of future events in the
physical worlds called laws of physics or chemistry?
-- Minimization of efforts towards prediction of human behavior as
single person / as an entire society?

AMong others. All good.


What comes in addition to my mind as aspects worth to be discussed in
this context are also:
- Does it in any area make sense to look for simplification instead
of acceptance that huge amount of data _is_ the way to go and
therefore it is better to concentrate on improving technology
for quick processing of terabytes of data than looking for a way
of reduction of amount of data to process?

Why not both? Diferent contexts to _start_ with.

- Does it make sense to cope with theories able to scale to infinity
(like e.g. Kolmogorov complexity) which results in controversy,
counter-intuitivity and other problems, instead of reducing any
problem to maybe a very huge but a finite amount of cases (e.g.
setting a given lower limit for possible resolution) where it is
always possible to find an answer to any reasonable question?

Actually only probably so in the first, and only possibly so in the
last.

Be sure when using the first that you do not include stuff you woud not
want to solve,
and in the last, you do not preclude the question you want the solution
to.


Claudio

[... critics about limits on Usenet ...]

The limiting problem about limits is that discussing them alone limits
the discussion.

Cheers.


From: Brian Raiter - view profile
Email: b...@xxxxxxxxxxxxxxxxxxxx (Brian Raiter)
Groups: comp.compression
Not yet rated
Rating:
show option

Reply | Reply to Author | Forward | Print | Individual Message | Show original | Report Abuse >| Find messages by this author


I am making a statement about the natural world. It is not just the
human brain that produces interesting non-random strings, but many
natural sources, such as DNA, image, video, all kinds of sensor data,
physical models, maps, etc.


Absolutely agree: Anthropocentric singel point of view blindness, or
in more extreme cases, arrognace or hubris, can result in loss of
this critical point of view. I am not certain I am absent of this
property
endocentrically myself: it is very hard to root out. If so.

Oftimes my cat (though he probably viewed it differently!) was much
more human
than most of the individuals I have interacted with on cerain topics,
or more
correctly, in his interactions with his "pet" (though liely,
cognitively, he
probably would not hav acknowleged another will within the common
noosphere we shard in reality that he probably woud not have recognized
as a universe distinct from himself: which was why he WAS (oh, this is
so
poorly expressed) the only operant will with the domain we shared:

did not matter, we got along OK, even though I was not really there.


Okay. Although (getting back to my original comment) it *is* the human


Sometimes. But perhaps your assertions hava blind spot: you have not
at al
theticized, much less proofed (as opposed to proved) the idea properly
that
ALL of it is so: just that at times i appears to be the case that
not all of it is NOT human oriented in that manner.

At least, IMOHO.


brain that picks through these natural data and selects the ones that
are "interesting". And, in most cases, it is the human brain that


Is it? Always? Hmmm. Varela may disagree with that. hard to say.


selects a representation of the information (say, sensor data) as a


sometimes it does: sometimes th selection does not appear o be able to
be ategorically discretely cleaved from any other operant metrifiable
property.

"Emergent" or "coalescent" situations defy this rather explicitly.


linear string of discrete data. So I would still say that human


It is not only discret data that are compresssible losslessly: only
that digital
systems have a damned hard time doing it. QED?


psychology is the main reason why most "interesting" strings are
eminently compressible.


ONe modal method among many: undoubtedly new, semi-based on semiotic
considerations on top of one or more flavors of proto-scientific
pseudo-psychology now coming into reality after being virtually present
for a while.

Useful, but not the only ay to recapitualte itl; and not always useful
in all contexts.


Of course, one could argue that the human brain is that way because it
was shaped by the same natural forces that made the other data
sources. Certainly I think that one could argue that highly complex


Context free langgaues are always either incomplete or inconsistent;
see Godel.

COntext sentive languages are not so: See Lesniewski.

And a language without a context has a problem: it cannot speak.

SO I am no too upset that perhaps a contet sentive language
is actually rqired to be in ore for mth to be consistent, while
requiring
something to talk ABOUT as its prerequisite.

Compression gets very intersting in this situation.


machines, such as living creatures, will naturally display a


I would prefer semiotic systems, not just complex machines.


preference for redundancy throughout their systems, and this could
easily result in the psychological biases I mentioned. This is a
little closer to what you were saying originally, I think ...


Sorry, circular, and not quite correct:

The presence of redundancyis a priori not used redundantly: it is usd
for
a constructive purpose you might call (lmiitedly and very
oversimplifiedly)
as error control: as such the redundancy present

is not redundant!

(chuckle, laugh: God is a
real corker at time; witnss the duck billed playpus being shovelled
into
the face of 18th century xenobiologists callig themselves scientists:
we learn every time we admit we have not learned yet. More or less:

(hiccup) 'Scuze me, I 'v been Taoing too much
)

Gotta go for now.

Your comments are not exactly wrong: they all have merit from a certain
POV;
but there is merit to examine the OTHER poits of view as well.

Why is a thing is it does not fill a particular function?

Cheers.

b



.

Relevant Pages

  • Re: Was: Re: What is the state-of-the-art analysing hardware impact on achievable compre
    ... As I had a very hard time to get through the text of the message below I hope, that eventual responses put some more structure into it making it easier to see the main points. ... Where comes the "redundancy" of data making compression possible from? ... What comes in addition to my mind as aspects worth to be discussed in this context are also: ... Compression gets very intersting in this situation. ...
    (comp.compression)
  • Re: Quick Huffman theory question
    ... application compresses and decompresses, compression appears to ... categories as context. ... good text compression requires gigabytes of memory. ... don't bother with Huffman codes. ...
    (comp.compression)
  • Re: Is the jpeg compression algorithm used in PAQ8h lossy?
    ... paq8h to get better predictions on the DCT coef values? ... next code using the decoded image as context. ... Stuffit transforms the jpeg image back to the DCT ... However, Stuffit gets better compression. ...
    (comp.compression)
  • Re: New large block BWT compressor
    ... could exploit full context information using strictly the BWT itself. ... if we think about the BWT string as order zero contexts as above we ... Here are some BBB compression results for block sizes from 10^1 through ...
    (comp.compression)
  • Re: Compression and crypto
    ... Actually if the compression scheme is well fit to the source ... For general-purpose use, the designer of the compression algorithm will ... a lot of the redundancy, but it won't eliminate all the redundancy. ... grammar or representing the fact that most plaintext is a semantically ...
    (sci.crypt)