Recipe improvement interface

Case number:699969-994438
Topic:General
Opened by:Roukess
Status:Open
Type:Suggestion
Opened on:Wednesday, January 30, 2013 - 17:26
Last modified:Wednesday, January 30, 2013 - 20:40

Hi everyone,

Most of us use and/or code recipes as they are powerful to solve puzzles. However, finding the best algorithm need many adjustments, just for the value of CI for instance, or even the algorithm by itself.

The variability of the puzzles doesn't permit to have an unbiased opinion of the efficiency of our new algorithms or settings. The size of the puzzles also slow down our recipes, and we sometimes have to wait a long time before the event of interest happens. Finally, as the aim of the game is to increase our points, it is difficult (or time consuming) to come back a the same starting point in order to compare the efficiency of each parameters we defined.

My suggestion :

Add a "recipe improvement interface".

It will permit to generate manually the protein we want (by writing the sequence) or to load already partially folded small proteins. It will also permit to create an other protein that we can "block", in order improve our protein-ligand recipes.

Points will only permit to compare the different algorithm we try, so here, no comparison between high scores.

What do you think about it? opinion? suggestion?

I know that it is a "big" suggestion, but Foldit has to give us everything we need to improve the solutions we submit.

(Wed, 01/30/2013 - 17:26  |  3 comments)


Joined: 06/17/2010

You can use contest to test scripts.

Joined: 10/26/2010

For sure, but in contest you will not have ED puzzles or ligand puzzles.

jflat06's picture
User offline. Last seen 1 day 7 hours ago. Offline
Joined: 09/29/2010
Groups: Window Group

I believe there are test puzzles for most types of puzzles in the devprev puzzle list. You can use these for testing purposes, though I realize that this will not allow you to compare devprev/main easily.

Sitemap

Developed by: UW Center for Game Science, UW Institute for Protein Design, Northeastern University, Vanderbilt University Meiler Lab, UC Davis
Supported by: DARPA, NSF, NIH, HHMI, Amazon, Microsoft, Adobe, RosettaCommons