Clarification of "outside" information

158 views
Skip to first unread message

moogie

unread,
Feb 24, 2020, 5:56:33 PM2/24/20
to Hutter Prize
There exists a rule: "The prize will be paid if the solution reflects the spirit of the contest. In particular decompressors (secretely) receiving any kind of "outside" information are forbidden. Also in order to verify your claim we need to be able to run your executable on our machines. Payment of the prize cannot be legally enforced. Marcus Hutter will make the final decision whether to recognize a record, award a prize, and the amount."

What constitutes outside information? 

Specifically, are we able to make use of information about the operating system? i.e. is this a restriction to only prevent querying information sources outside of the test environment?

I want to attempt a create a submission where tailoring can be performed based on the information provided by the operating system in lieu of a common dictionary between the compessor and decompressor.

Matt Mahoney

unread,
Feb 25, 2020, 11:15:06 AM2/25/20
to Hutter-Prize
Standard libraries are OK. Nonstandard files are not. For example, you can't use /use/dict/words because it's not standard in Windows.

--
You received this message because you are subscribed to the Google Groups "Hutter Prize" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hutter-prize...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hutter-prize/52f74542-1d80-4973-b0a6-51d5aed28788%40googlegroups.com.

moogie

unread,
Feb 25, 2020, 5:24:39 PM2/25/20
to Hutter Prize
Yes, i agree, expecting files to be present is not going to be viable (e.g. /use/dict/words in windows) instead I plan to have the algorithm OS agnostic:

To attempt to remove the need for a common dictionary present in both compressor and decompressor, I going to have the program perform a search for available, text files as provided by the OS default install, to train the internal model with. Obviously this does make the resultant compressed output to be highly coupled to the computer that produced it, and the compression ratio will differ based on the files found (i.e. between operating systems) however for the purposes of the competition it seems an acceptable restriction.



James Bowery

unread,
Feb 25, 2020, 5:26:54 PM2/25/20
to Hutter Prize
That is a clear violation of the spirit of the competition.

On Tue, Feb 25, 2020 at 4:24 PM moogie <budge...@mystarship.com> wrote:
Yes, i agree, expecting files to be present is not going to be viable (e.g. /use/dict/words in windows) instead I plan to have the algorithm OS agnostic:

To attempt to remove the need for a common dictionary present in both compressor and decompressor, I going to have the program perform a search for available, text files as provided by the OS default install, to train the internal model with. Obviously this does make the resultant compressed output to be highly coupled to the computer that produced it, and the compression ratio will differ based on the files found (i.e. between operating systems) however for the purposes of the competition it seems an acceptable restriction.



--
You received this message because you are subscribed to the Google Groups "Hutter Prize" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hutter-prize...@googlegroups.com.

moogie

unread,
Feb 25, 2020, 8:25:07 PM2/25/20
to Hutter Prize
I was wondering if that was the case. I guess i was rationalising it as treating it similar to how most (all??) living learning beings that come into world with core rudimentary knowledge (instinct?) on which they then use to then expand upon.

If it is in violation then that is fine. I will not persue that avenue.

I just wonder where the line is drawn? Matt suggests that standard libraries (i guess STL?) is fine, but isnt that in its self knowledge? e.g. not having to implement space-expensive functions within your compessor/decompressor. How is using files present in the OS any different?


On Wednesday, 26 February 2020 08:56:54 UTC+10:30, James Bowery wrote:
That is a clear violation of the spirit of the competition.

On Tue, Feb 25, 2020 at 4:24 PM moogie <budge...@mystarship.com> wrote:
Yes, i agree, expecting files to be present is not going to be viable (e.g. /use/dict/words in windows) instead I plan to have the algorithm OS agnostic:

To attempt to remove the need for a common dictionary present in both compressor and decompressor, I going to have the program perform a search for available, text files as provided by the OS default install, to train the internal model with. Obviously this does make the resultant compressed output to be highly coupled to the computer that produced it, and the compression ratio will differ based on the files found (i.e. between operating systems) however for the purposes of the competition it seems an acceptable restriction.



--
You received this message because you are subscribed to the Google Groups "Hutter Prize" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hutter...@googlegroups.com.

James Bowery

unread,
Feb 25, 2020, 8:31:17 PM2/25/20
to Hutter Prize
Read "standard OS libraries" as "standard instruction set".

One can play games with words but all you'll end up doing is wasting your time writing an algorithm that will be rejected by everyone else -- not just the judges.

Here's another example of a word game:

"What do you mean 'instruction set'?  Couldn't one have an 'instruction' that is like an 'oracle' and delivers 'outside knowledge'?"

This is the last word on this.  

To unsubscribe from this group and stop receiving emails from it, send an email to hutter-prize...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hutter-prize/d315b7ea-bed7-4ce2-b501-d87e4e2ddb6f%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages