I've thought of a data compression technique which I'd like someone to
implement in software.

The compression technique is described at:
http://www.halfbakery.com/idea/Sliced_20cube_20data_20compression#1136339984

Preferably this would be coded in Matlab (or equivalent such as Octave).

Any comments on the originality or practicality are welcome.

---

Request for Question Clarification by hedgie-ga on 09 Jul 2006 00:14 PDT

It looks like you do not want any answers, just comments.
That would be an abuse of this service.

 If I am wrong, and you want an answer 
( which would be
 'it is original and totally impractical'
 and some references to books and papers on compression and .. )

please explain what you expect from the answer. 
Apparently, you do not want to have it patented,
so what is your motivation and purpose?

---

Clarification of Question by xaviergisz-ga on 15 Jul 2006 04:15 PDT

Hi hedgie,

I was hoping someone with coding experience would be able to implement
this so I could experiment with it.

but I would accept as an answer why this is 'it is original and
totally impractical'

Another answer that I would accept would be what types of data this
would suit. I realise that any lossless data compression will only be
useful for limited types of data (by the pigeon-hole principle).

sorry for the lateness of this reply.

---

Request for Question Clarification by hedgie-ga on 15 Jul 2006 16:49 PDT

hi  xaviergisz-ga


1) sorry for the lateness of this reply.

    That's OK . We are all frustrated because email notifications are
not working properly lately

2) I did re-read and re-read your description,
 did a short search for references and decided to just post a comment
 - which means : no cost to you and no obligation to clarifications  my part.

  The comment describes my impression and gives few references I looked at.
  It is more a 'review' then an answer.

Hedgie

Opinion

  The compression algorithm transforms elements selected from some emsamble of
  data-sets (texts, images, ..).
  It works well if there is match between the ensemble and transformation.

  Here, we have a description of transformation - and a question : "On which
  data sets it will work well?" - sort of the Inverse Problem to the usual
  approach (which takes ensambles from applications and looks for 'suitable
  encodings').
 
  The algorithm is described intuitively, 
  by geometrical transformation of 2D images stacked to form 3D structure.
  If we do the slicing 'cleverly' and 'data are nice' we get a good compression.

  Problem of implementing that programmer is really not told what is a clever,
  what geometrical manipulation is good for what ensembles. It restates the
  concept of encoding and gives only an example of transformation for one ensemble.

  I would suggest to describe general algorithm in pseudocode  

SEARCH TERMS:  what is pseudocode

so that programmer knows what 'clever' and 'nice' mean (for a given
ensemble types?)

  Here a few references, overviews, and a textbook, showing that that this sketched
  idea is among the main techniques used in intros. It may be
original, however full
  literature search was not done. That, as well as coding (be it in c
[preferable] or
  high level matrix language, such as Octave, would IMHO far exceed
the offered price.

 references:
  
  
 Introduction / Lossless Data Compression - basic terms and methods


http://www.vectorsite.net/ttdcmp1.html
http://www.data-compression.com/lossless.shtml


DATA COMPRESSION STATISTICS AND IMPLICATIONS
Many lossless data compression algorithms exist. Some of the main
techniques in use are the Huffman [4], Arithmetic [3], Lempel-Ziv [5],
runlength, predictive coding or variations and combinations of these.
Each of these methods can be found in most data compression texts ..
http://www.ece.nmsu.edu/~sheila/ITC.htm


 The Design and Analysis of Efficient Lossless Data Compression
Systems. PhD thesis, Department of Computer Science, Brown University,
June 1993.
references

http://citeseer.ist.psu.edu/context/80210/12854


textbook
By David Salomon. Published by Springer (2000). ISBN 0-387-95045-1.
LCCN QA76.9.D33 S25 2000. xvi + 823 page

http://www.ecs.csun.edu/~dsalomon/DC2advertis/DComp2Ad.html

hi hedgie,

thanks for your thoughts on this somewhat peculiar question.

stacking the data 'cleverly' means for a given data set compressing it
*efficiently*, or even more preferably, compressing it *optimally*
using this method.

I was hoping to discover myself the 'nice' (ie compressible) types of
data using the code that someone could write. Although I thought it
might be obvious to someone what data would be compressible without
the need for the software.

I thought the price offered was not unreasonable, since the problem
being solved is a more of an interesting puzzle than a labourious
exercise, and I imagine it could be achieved in about 20 lines of
code. I would happily increase the price if anyone wanted.


Going back to your original clarification: 'Apparently, you do not
want to have it patented, so what is your motivation and purpose?' I'm
just curious (but have not ruled out the possibility of patenting
(keep in mind the 'grace period' of 12 months between publication and
application).