4 replies [Last post]
Joined: 05/19/2009
Groups: Go Science

Is it possible to use function such as math.abs and math.random in the cookbook?
If so, how do I import them?

Joined: 11/10/2007
Groups: Window Group
They are not currently

They are not currently available, however, that functionality may be added in the future.

Joined: 09/18/2009
Groups: SETI.Germany
Some help...

As long, as they are not available, I can give you a self-written absolute-function:

function abs(x)
if y <0 then y=-y end
return y

Do you also want an integer function?

Joined: 05/19/2009
Groups: Go Science
In your function you don't

In your function you don't need to different variables.
function abs(y)
if y <0 then
return -y
return y

If you have a nice random function I would be interested.
The randomness is probably best produced at the moment by taking the last digits of the score through.

If you have a good ideas/code on doing split testing I would also be grateful. At the moment I don't have a way to test whether a change improves the recipe or doesn't.

Joined: 09/18/2009
Groups: SETI.Germany
Random numbers

Yes, in our (TheMarquis and me) random bond recipe, we use the sum of segment-scores as seed value, multiply it with a high value to get some "noise" and then perform a modulo operation.

And the way how you mix the segment scores together and multiply them will give some good variations:


To check, if a recipe changed the score in a positive way, I store the sum of all segment-scores before (in this way you can also grab scores below zero) and do "save recent best".
Then I perform my desired action.
When the recipe has finished, I calculate the score-sum again,
and compare this score-sum result with the score-sum of the initial state.
If the initial state is better than the new state, I do "restore", else i keep the new puzzle result.

Also, have a look here:



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