Hutter Prize Google Group

Re: [Hutter Prize] Stochastic Context Free Grammar

2008-09-19T14:20:21Z

I'm not aware of PCFG being applied to text compression, although in theory it might work. The top compressors like paq8hp12 and durilca4linux model language mostly at the level of n-grams and a crude semantic/part-of-speech model in which related words are grouped in a predefined dictionary and whole groups are used as context. I believe that durilca4linux used a clustering algorithm to find related words. However, its top position in the LTCB is probably because it uses twice as much memory. The model itself is more primitive because it uses PPM rather than context mixing, which restricts the semantic model to a 1-dimensional space.

Stochastic Context Free Grammar

2008-09-19T12:21:24Z

Hi!

I've only found the Hutter prize some days ago...:)
I am wondering: did anyone try an approach of creating an algorithm,
which automatically learns a Stochastic Context Free Grammar (
[link] ) based on the data?

Creating such an algorithm seems to be non-trivial, so before working

Ocarina Announces $1 Million In Compression Prizes

2008-09-11T05:54:06Z

OCARINA UNVEILS $1 MILLION PRIZE FUND TO ADVANCE THE STATE OF
COMPRESSION RESEARCH FOR DATA STORAGE

Prize Announcement Follows Compression Summit Featuring Leading
Industry Researchers

September 10, 2008 – San Jose, Calif. – Ocarina Networks, provider of
the industry’s first online storage optimization solution, today

Re: The XML content

2007-10-13T15:19:23Z

Hi!
It is incorrect. You can try next XML tricks:
1) replacing <a>...<b>...</b>...</a> with
<a>...<b>...SpecSymbol...SpecS ymbol by stack machine;
2) moving contents of <timestamp>, <id>, <revision>, <page> to the
separate stream;
3) moving contents of <text>, <title>, <comment> to the separate
stream;

The XML content

2007-10-11T17:34:14Z

I'd like to chime in with some interesting empirical results from my
attempts to compete for this prize. In due time I'll reveal all my
successful techniques, but I wanted to contribute to the discussion
here with a fascinating and revealing unsuccessful technique.
Throughout the history of the contest, there has been a latent

Re: [Hutter Prize] Re: Verified HKCC-2

2007-09-02T10:06:33Z

I verified the Sept. 1, 2007 update of HKCC-2.
16,234,075 compressed_enwik8.dat
14,420 decompress.exe
16,248,495 bytes
Decompression time was 11847 sec. (Athlon-64 3500+, 2 GB, WinXP) using 880 MB
memory slowly climbing to about 940 MB. Output matched enwik8.
decompress3.exe does not appear to be updated for the new compressed file, but

Re: [Hutter Prize] Verified HKCC-2

2007-09-01T20:03:48Z

I verified decompress3.exe on a 1.7GHz Athlon, running in 5 hr 45 min.

Verified HKCC-2

2007-08-31T14:55:28Z

I have verified Alexander Ratushnyak's entry as of Aug. 29 2007.
[link]
The program is a decompressor and a file that decompresses to enwik8.
I used the command
decompress compressed_enwik8.dat enwik8
On an AMD Athlon-64 3500+, 2 GB memory, Windows XP SP2 (32 bit) the
program output enwik8 (verified identical) in 11802 seconds process

Re: Alexander Ratushnyak Wins Second Hutter Prize Payout

2007-07-24T18:06:59Z

I should have done this earlier, but paq8hp12_any source code is now
posted to [link]
I also posted a "lite" version of paq, lpaq1, that might be useful for
others entering the contest. I wanted an open source context mixing
compressor that's small and fast enough to build on. It is 35 times

Alexander Ratushnyak Wins Second Hutter Prize Payout

2007-07-09T22:59:53Z

Alexander Ratushnyak <[link]> is the
second winner of The Hutter Prize for Lossless Compression of Human
Knowledge <[link]>. The Hutter Prize was
established and initially underwritten by Marcus Hutter <http://
en.wikipedia.org/wiki/Marcus_H utter> based on the equivalence between

Re: paq8hp11

2007-05-16T22:12:25Z

On May 14 I confirmed compression and decompression for paq8hp12 -7
enwik8.
[link]
16,381,959 enwik8.paq8hp12
99,696 PAQ8HP12.EXE
16,481,655 total size
Timer 3.01 process time was 13082 sec. to compress, 13148 sec. to
decompress
on a 2.2 GHz Athlon 64 3500+ with 2 GB memory. Windows Task Manager

Re: [Hutter Prize] Re: paq8hp11

2007-05-10T12:07:07Z

No, not yet. I was a bit busy, but now that paq8hp11any is out, i can
compress it with 512 Mb. Thanks anyway.

Re: paq8hp11

2007-05-04T17:30:34Z

Oh, sorry, I deleted it after the test. Do you already have the
compressed file? It should be 16,459,515 bytes.
-- Matt Mahoney

Re: paq8hp11

2007-05-04T13:50:10Z

Could someone please send me the md5 (or other hash) of the compressed
archive? I don't have enough memory, but I'll try to do it with
virtual memory, probably under wine environment, and I want to be sure
I have the right file before trying to uncompress (it will be a time
consuming process). I hope I'm not asking too much...

Re: [Hutter Prize] paq8hp11

2007-05-01T22:34:01Z

My results were the same for an AMD Athlon Xp 2100+ 1.73 GHz with 1.25
GB of RAM except it took compression and decompression took almost
exactly 6 hours.