Re: The Prize #2

[Chris Albertson <chrisalbertson90278@yahoo.com>]

--- tom <tdroege2@earthlink.net> wrote:

> The response to the first prize offer was pretty underwhelming
> considering the 
> size of the prize offered.

You need to get the word out to a group larger than the TASS
mailing list.  University students in Computer Science and
Astonomy.

Get people from "outside" to work on this.  

Make a big and atractive webpage the describe the prive and
the problem in terms someone with some education but knowing
Nothing about TASS could read.

You've got a Wiki.  start a page about possable methods to
improve the data there.  

You may want to allow group efforts by teams

Some ideas about the rules.  
1) Limiting the hardware is a good idea.  If it requires a half year
on a 100 node cluster you can't use the result.  But don't put
to many restriction on it.  Maybe you can raise the prize to
$25K and then say the prize will be reduced by the cost of the
hardware and any other equipment  Or maybe it's a $25 prize and the
winner has to hand over the hardware?
2) The tools must be freely available and all source code must be
provided.
3) The result must by replicated by (at least) one other party
using the tools
provided in #2 above. (This insures that (a) ALL the tools where
in fact delivered and (b) the data was not "cooked".

A $20K prize is large enough that someone might cheat.  I could
get the scatter in the results down to << .0001 by simply
writing a program that determines that ALL stars have magnitude 5.0
"scater" should not be the only criteria.

OK if making them all 5.0 is disallowed what if I round each
magnitude to the nearist 0.1 mag.  That's "bining" and it would
reduce scatter too.  What about other more clever ways to
cheat that are harder to detect?  How about rounding to a
direction determined by crytographic
hash function of the RA, DEC location?  A good crypographic
hash would be statistically undetectable but could reduce
statter to near zero.

I'm pretty sure I could "improve" the TASS data by any factor
with just a simple hack that "fudges" the result of each star
by no more then 0.05 mag.  If the code to implement the hack
where distributed, one line here and another line there over
a 20,000 line program you would not find it.

It will be hard to come up with a definition of what it takes to
win.  You would need to know the magnitude to every star in
advance but then if you did why do a survey?

I once entered a bridge building contest and got third place
The winner read the rules more carfully then I did and found
that while you could use only 40 linear feet of 1/8" balsa wood
the amount of glue was not limited.  He build his bridge out
of dried glue wih a bit of balsa wood in it.
   

> My first assumption is that Americans
> have lots 
> of money and little time.  Not so true in Canada where people seem to
> have 
> more time and less money.  The statistics are not very good.  ;^)  I
> had said 
> (all mail lost) that I wanted at least a half dozen respondents to go
> through 
> the work of investigating the tax and other consequences.  I got 2 or
> three.   
> Only two really responded in a positive way.
> 
> Still, the quality of those that did apply is exceptional.  So I am
> quite 
> tempted.  I just hope a few more will show interest.
> 
> For those that missed he first mailing, I have proposed a prize of
> $20,000 for 
> work leading to the improvement of the tass data.  I think it would
> be 
> appropriate to have a debate here as to what the rules should be, and
> then we 
> would set a fairly long period to demonstrate a positive result.  Say
> the 
> year 2005.  I think pushing the noise floor below 0.02 or 0.01 would
> be a 
> suitable goal for such a prize.  No result, no prize.  The method
> would also 
> have to be practical.  I have to be able to process all the data in a
> year 
> using a half dozen or so GHz computers. 
> 
> Jennifer exploded when she saw the first offer but I think I have
> convinced 
> her that this is a good way to advance a project that I have already
> spent 
> much more in pursuing.  
> 
> If any of you are interested in participating, send again a short
> mail 
> message.  For those of you new to me it should include some
> indication of 
> your qualifications.  
> 
> 


=====
Chris Albertson
  Home:   310-376-1029  chrisalbertson90278@yahoo.com
  Cell:   310-990-7550
  Office: 310-336-5189  Christopher.J.Albertson@aero.org
  KG6OMK


	
		
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - 100MB free storage!
http://promotions.yahoo.com/new_mail