Re: QI and MQ Coder: First real-life experiences

nightlight wrote:
You are correct. I see why now his numbers are so small.
I forget he has to have all these exra fields around for the
side information.

I always compare side information according to what the other
coder does with it. The Moffat98 includes lengths of data in its
output, hence QI's figures include that cost as well when comparing
to Moffat98 AC, as my original post on N=10^6, A=3 did, where I

Actually you don't do the comparsion in a fair why which is
way one should do the test with files. Something you can't
seem to do. My guess is you can't compress as small even
for this simple task.

compared QI's output to their output. Since they also use adaptive
mode, which is not necessary and hurts their result here, I even
included that cost in the comparison (the 39 bits estimate for
frequency table). You can see all that in the original post which
reported those results:

http://groups.google.com/group/comp.compression/msg/ff1ee67d18b63f5a

But if your coder doesn't include N and A into the compressed
output (but relies instead on OS to remember N via file size
and on hardwired context tables & Huffman coder to remember
alphabet size A), then I don't add those figures to the QI's output
size either. It is you who kept repeatedly quoting my figures
stated explicitly for different coding conditions than yours (where
N and A are not given to the decoder as parameters but must
be included into the compressed data) so you can claim
to do better.


First of all I showed how a length field could be added
so it could be used in a buffer it seems you forgot this but
it was added in even in a nonoptimal way and it beat your
code. I think its best of any one cares to read the previous
posts. And I am sure you know this even though you claim
that N was not accounted for. We could play this game
forever. That why people in this field like code they can use
themselves with files of there choice. What is it about
allowing people to do real test with files that seems to
scare you. What are you really afraid of.? Do you think
more people will laugh at you if they see it fail or do you
only feel safe saying what it can do with out actually doing
it or allowing others the choice of using there own test files.


Yes I kept quoting your figures since mine where better
you didn't seem to understand why I compressed the pure
case of all A's longer its still not clear you have a concept
of real arithmetic compression. Now all the sudden you
come up with different numbers. Friend thats why people
do it with files so others can check. At this point do you
really think anyone cares what you currently claim. Can
you get it to work wifh files. I guess not. Will you code
it to files so others can fix it I guess not. WHY?

Is QI a drop in coder for arithmetic when used as an
entropy coder yes or no? Isn't QI more flexable
yes or no?

Either you can't do it or you have and now that you failed
after all this was what you claimed was a fair test. I guess
now that you can't do or wont do it. You have to come
up with some other reason. You don't have a concept
what the conditions are. Your mind will not currently
allow the results to show your wrong. Again that is why it
has to be to files so all can check. Is that so hard to
understand?

And now, after blowing the smoke for several threads, conveniently
mixing up the coding results from different coding conventions
throughout (despite being explicitly reminded numerous times of
what each figure represents), you declare your sincere concern
that it was me who is using such infantile gimmicks. Whoah Nelly.

Whoah Nelly is right. Who is mixing coding results. Good question
how can it be resloved. You could post hundreds of line of text or
you could write a program that actaully allows people to test files
and see what happens. Let me guess you will whine and write theory
as to why you are correct but you will not write code. Please prove me
wrong with real code. That is if QI can actually do what you claim
it can do.

It seems to me you aren't really interested in comparing the
actual results of a compression. I suspect it would not be hard
to write a pure arithemtic that only calculates the index. Which
it seems is the only thing QI can do. I suspect its actually easier
to code the index using arithmetic in the first place. I suspect why
this is not done very often for compression is due to the extra side
information you have to carry with the index to do any real
compression.
Which may be the main reason its so hard to use the index for
a drop-in to replacement for what is best done by a pure arithmetic
coder
in the first place.


Maybe if you ever get working code for this one simple example
we can compare results in just calculating the index only. A proper
arithmetic should be able to calculate the index as a n and n+1
bit string. Of course I am sure you can quote all kinds of theory
to show this is not possible. So that even for this you would not
risk a real code test. Just quoting text and saying things are so is
more than enough for you. But here at this forum we get people
all the time quoting stuff with out real code.


David A. Scott
--
My Crypto code
http://bijective.dogma.net/crypto/scott19u.zip
http://www.jim.com/jamesd/Kong/scott19u.zip old version
My Compression code http://bijective.dogma.net/
**TO EMAIL ME drop the roman "five" **
Disclaimer:I am in no way responsible for any of the statements
made in the above text. For all I know I might be drugged.
As a famous person once said "any cryptograhic
system is only as strong as its weakest link"

.

Relevant Pages

  • Re: QI and MQ Coder: First real-life experiences
    ... I always compare side information according to what the other ... be and very much in line with my AC coder. ... When the encodeing is through, ... stated explicitly for different coding conditions than yours (where ...
    (comp.compression)
  • Re: New (and free) DICOM Compare tool
    ... Speaking of compression, I am still stumped by an old Siemens format. ... read the image's header O.K. but the pixel data is compressed. ... run it through an in-house "DICOM Compare" to compare the conversion ... If you drop 1 DICOM file on the program, it will compare that file to ...
    (comp.protocols.dicom)
  • Re: QI and MQ Coder: First real-life experiences
    ... I always compare side information according to what the other ... be and very much in line with my AC coder. ... When the encodeing is through, ... stated explicitly for different coding conditions than yours (where ...
    (comp.compression)
  • Re: Pointers for JPEG lossless compression verification ?
    ... Tools for comparing images for lossless compression are as easy as you ... said: Compare all pixels. ... source and target, i.e. PNM, then a suitable tool to check for ...
    (comp.compression)
  • Re: really nice preamp
    ... If you want to compare the FMR stuff to Focusrite, ... compression in your stated application will do more harm than good. ... you can get two channels of really ... about RNP for my studio work. ...
    (rec.audio.pro)