[freeciv-ai] (PR#10203) Greedy CM algorithm

["Jason Short" <jdorje@xxxxxxxxxxxxxxxxxxxxx>]

<URL: http://rt.freeciv.org/Ticket/Display.html?id=10203 >

> [bhudson - Fri Nov 19 04:57:14 2004]:
> 
> On Thu, Nov 18, 2004 at 06:07:30PM -0800, Jason Short wrote:
> >
> > <URL: http://rt.freeciv.org/Ticket/Display.html?id=10203 >
> >
> > On large cities the B&B algorithm performs much worse than the DP
> algorithm.
> >
> > I'm only talking about the client here.
> >
> > For instance in the game at
> > http://rt.freeciv.org/Ticket/Attachment/75375/51725/jason-
> game.sav.gz
> > (which is actually a real game of mine, unlike all the made-up games
> > I've been testing with) the DP algorithm takes 0.1 seconds to
> > recalculate Boston while the B&B algo takes 0.8 seconds.  Not good!
> 
> Really?  I'm surprised, because usually large cities have stringent
> demands on food, which means the vast majority of solutions the DP
> algorithm dreams up are pruned out.

In this case the food requirements are stringent.  Every available
non-specialist tile must be used for food production.  But since there
are many citizens left over these can be put into any specialist role. 
And there is no particular method for pruning them.  Yet they should be
easy to prune, since according to my weights elvises should add 0
fitness, taxmen 8 fitness, and scientists 14 fitness.

> By the way, what's the CMA settings?  The gui (gtk2) isn't very good
> about custom CMA settings: when I switch to the CMA tab, it decides to
> apply the first setting in my list.  Which makes it harder to debug.

20/5/0/1/1/0 I think.

-jason