Page 1 of 1

Automatic evaluation

PostPosted: Sun Nov 23, 2008 3:17 pm
by guest2
It's good that CASP8 will release an "official" automatic evaluation. There are
serveral reasons that the server section should play a more important role
than the human section this time (or at least should be paid more attention
to than previous CASPs). First, there are more registered server groups than
the human groups for the first time in CASP8 (121 vs. 112); second, there
are more server targets than the human ones this time; third, to some extent,
the results of automated server predictions reflect more objectively the
advance of the algorithms than the human-intervened results; fourth, maybe
most importantly, only the advancement of the automated predictions (rather
than the human-expert predictions) could be exploited by the biology community
(through on-line servers or software release), which has been THE goal of the
CASP experiments.

Someone may add more reasons to the list. :)

Re: Automatic evaluation

PostPosted: Thu Nov 27, 2008 1:45 pm
by djognson
Cannot agree more. Do we really need a human section in CASP9
if it turns out to be dominated by ranking/selecting server models
or by reading related mutagenesis papers? Combining and refining
server models to generate better ones may be one valued point
for human section. But it can also be done through server section
and the result is to me more trustable.

Re: Automatic evaluation

PostPosted: Thu Sep 16, 2010 12:31 am
by AlexanderSimmons
It will be provided with the results of numerical forecasting evaluations performed at the Prediction Center, and will evaluate the results primarily on that basis. They will be asked to focus especially on the strength of different methods.