| Claudio Grondi 2005-11-15, 7:55 am |
| Thank you for your reply opening my blind eyes.
> The example that John_H posted had this form.
Trying to understand why I had not seen it myself I found, that I was
 by the comment "I just added an extra zero to the 7-bit 0 sequence
to make it all 256" assuming, that "the first 7 bits of the sequence
appended to the end" were the first 7 bit of the genuine "maximal length
sequence" and not the actual sequence posted, so I have overseen that the
last byte of the posted sequence is the same as its first.
> But your dream of exploiting this for compression is doomed to failure.
Trials to compress the 256 bytes data string generated from the bits of the
"maximal length sequence" with bzip2, zip or lzma compression scheme failed,
so compressing it down to 32 byte with the 'window' approach can't be seen
as a failure like it is also the case with all pseudo random number
sequences where the generator program itself is the best compression scheme
available.
I have to admit, that the value of this in practical cases in which
compression programs are commonly used probably approaches zero, but it does
not make it less interesting for the special cases in which it works and as
source of wisdom when analysed deeply in the attempt to gain some more
understanding of what compression actually is.
I like the idea of a 'window' floating above a data pattern much more than
the idea of a 'linear feedback shift register' where it is much harder for
me to 'see' and understand what it actually does. If it should turn out,
that both approaches lead to same results, I would prefere to talk about it
in terms of a 'window' than in terms of 'shifting registers with linear
feedback'.
I will probably not dig deeper into it now, but there are already pictures
of 2D and 3D 'window's floating above 2D and 3D data patterns in my mind.
Claudio
"Ben Rudiak-Gould" <br276deleteme@cam.ac.uk> schrieb im Newsbeitrag
news:dlcgb3$7jm$1@gemini.csx.cam.ac.uk...
> Claudio Grondi wrote:
down to[color=darkred]
to[color=darkred]
>
> The example that John_H posted had this form. He even included a comma to
> show where it wraps around.
>
coming[color=darkred]
power[color=darkred]
>
> Such sequences can be generated using a linear feedback shift register
> (which google).
>
> But your dream of exploiting this for compression is doomed to failure.
>
> -- Ben
|