Thanks a lot.
Indeed they are, it is a nice mailing list. Actually I got from there the
background of the typed GP concepts.
Enjoy Rome.
On Thu, Dec 17, 2009 at 1:52 PM, Sean Luke <[log in to unmask]> wrote:
> It's good to see you got it working. Sorry I couldn't respond in time --
> I'm presently in Rome this semester and my turnaround is much slower than
> normal.
>
> I used to have a variant of Artificial Ant which did typed GP for a demo
> example, but sadly I can't find it anywhere. FWIW, if you search the ECJ
> mailing list archives you'll also find a lot of examples and discussion
> about strongly typed GP.
>
> Sean
>
>
> On Dec 17, 2009, at 12:11 PM, alibaba wrote:
>
> Hi all again.
>>
>> I added the restrinctions that i needed (see bellow). I this thread will
>> be useful to somebody.
>>
>> Regards.
>>
>>
>> gp.nc.size = 16
>>
>> gp.nc.7 = ec.gp.GPNodeConstraints
>> gp.nc.7.name = ncAND_or_rule
>> gp.nc.7.returns = and
>> gp.nc.7.size = 2
>> gp.nc.7.child.0 = or
>> gp.nc.7.child.1 = rule
>>
>> gp.nc.8 = ec.gp.GPNodeConstraints
>> gp.nc.8.name = ncAND_and_rule
>> gp.nc.8.returns = and
>> gp.nc.8.size = 2
>> gp.nc.8.child.0 = and
>> gp.nc.8.child.1 = rule
>>
>> gp.nc.9 = ec.gp.GPNodeConstraints
>> gp.nc.9.name = ncAND_rule_rule
>> gp.nc.9.returns = and
>> gp.nc.9.size = 2
>> gp.nc.9.child.0 = rule
>> gp.nc.9.child.1 = rule
>>
>> gp.nc.10 = ec.gp.GPNodeConstraints
>> gp.nc.10.name = ncOR_or_rule
>> gp.nc.10.returns = or
>> gp.nc.10.size = 2
>> gp.nc.10.child.0 = or
>> gp.nc.10.child.1 = rule
>>
>> gp.nc.11 = ec.gp.GPNodeConstraints
>> gp.nc.11.name = ncOR_and_rule
>> gp.nc.11.returns = or
>> gp.nc.11.size = 2
>> gp.nc.11.child.0 = and
>> gp.nc.11.child.1 = rule
>>
>> ##Not used
>> #gp.nc.12 = ec.gp.GPNodeConstraints
>> #gp.nc.12.name = ncOR_rule_rule
>> #gp.nc.12.returns = or
>> #gp.nc.12.size = 2
>> #gp.nc.12.child.0 = rule
>> #gp.nc.12.child.1 = rule
>>
>> gp.nc.13 = ec.gp.GPNodeConstraints
>> gp.nc.13.name = ncAND_and_and
>> gp.nc.13.returns = and
>> gp.nc.13.size = 2
>> gp.nc.13.child.0 = and
>> gp.nc.13.child.1 = and
>>
>> ##Not used
>> #gp.nc.14 = ec.gp.GPNodeConstraints
>> #gp.nc.14.name = ncOR_and_and
>> #gp.nc.14.returns = or
>> #gp.nc.14.size = 2
>> #gp.nc.14.child.0 = and
>> #gp.nc.14.child.1 = and
>>
>> gp.nc.15 = ec.gp.GPNodeConstraints
>> gp.nc.15.name = ncRule
>> gp.nc.15.returns = rule
>> gp.nc.15.size = 0
>>
>>
>> gp.fs.0.size = 7
>>
>> gp.fs.0.func.0 = ec.app.gp.ossimRuleGeneration.func.And
>> gp.fs.0.func.0.nc = ncAND_or_rule
>>
>> gp.fs.0.func.1 = ec.app.gp.ossimRuleGeneration.func.And
>> gp.fs.0.func.1.nc = ncAND_and_rule
>>
>> gp.fs.0.func.2 = ec.app.gp.ossimRuleGeneration.func.And
>> gp.fs.0.func.2.nc = ncAND_rule_rule
>>
>> gp.fs.0.func.3 = ec.app.gp.ossimRuleGeneration.func.Or
>> gp.fs.0.func.3.nc = ncOR_or_rule
>>
>> gp.fs.0.func.4 = ec.app.gp.ossimRuleGeneration.func.Or
>> gp.fs.0.func.4.nc = ncOR_and_rule
>>
>> gp.fs.0.func.5 = ec.app.gp.ossimRuleGeneration.func.Or
>> gp.fs.0.func.5.nc = ncOR_rule_rule
>>
>> gp.fs.0.func.6 = ec.app.gp.ossimRuleGeneration.func.rule
>> gp.fs.0.func.6.nc = ncRule
>>
>>
>>
>>
>>
>>
>> On Tue, Dec 15, 2009 at 5:32 PM, alibaba <[log in to unmask]> wrote:
>> Hi All,
>>
>> Thanks for keep on reading.
>>
>> I have found what I was looking for on this thread: The "nil" GPType in
>> GPNodeConstraints. I am indeed trying to add the constraints that my model
>> needs and I'm coming across some troubles.
>>
>> Let me contextualize the problem. I'm generating a tree of AND and OR
>> non-terminal nodes and a RULE-function as terminal nodes. I want to add some
>> constrains so the generated tree follows this two rules:
>> 1) The root of the tree must be an AND function.
>> 2) At least one of the childs of the non-terminal nodes must be or a
>> terminal-node (RULE) or an AND node.
>>
>> I'm trying to add the described constrains like this:
>>
>> First, I define the GPTypes as follows:
>>
>> gp.type.a.size = 4 gp.type.a.0.name = nil # This one might not be
>> necessary gp.type.a.1.name = rule gp.type.a.2.name = and gp.type.a.3.name= or gp.type.s.size = 0
>>
>> To fulfill 1) I added the follow tree constrain:
>>
>> gp.tc.0.returns = and
>>
>> To fulfill 2) I added the follow node constrains:
>>
>> gp.nc.size = 10
>> #--------------------# gp.nc.7 = ec.gp.GPNodeConstraints gp.nc.7.name =
>> ncAND gp.nc.7.returns = and gp.nc.7.size = 2 gp.nc.7.child.0 = rule
>> gp.nc.7.child.1 = or
>> #--------------------# gp.nc.8 = ec.gp.GPNodeConstraints gp.nc.8.name =
>> ncRULE gp.nc.8.returns = rule gp.nc.8.size = 0
>> #--------------------#
>> gp.nc.9 = ec.gp.GPNodeConstraints gp.nc.9.name = ncOR gp.nc.9.returns =
>> or gp.nc.9.size = 2 gp.nc.9.child.0 = rule gp.nc.9.child.1 = and
>>
>>
>> Then I add each constrains to the functions of the function set:
>>
>> gp.fs.0.func.0 = ec.app.gp.ossimRuleGeneration.func.Or gp.fs.0.func.0.nc= ncOR
>> gp.fs.0.func.1 = ec.app.gp.ossimRuleGeneration.func.rule gp.fs.
>> 0.func.1.nc = ncRULE
>> gp.fs.0.func.2 = ec.app.gp.ossimRuleGeneration.func.And gp.fs.0.func.2.nc= ncAND
>>
>>
>> Note that with this constrains I am being more restrictive that what I
>> have defined in 2), and actually it doesn't work because, as it is obvious,
>> it never finish due there is always a non-terminal node as child in
>> gp.nc.8.child.1 and gp.nc.9.child.1. The problem here is that I need
>> gp.nc.8.child.1 and gp.nc.9.child.1 to be some times AND and some others
>> RULE, so the GP program will eventually finish.
>>
>> And I will like to set gp.nc.8.child.0 and gp.nc.9.child.0 some times to
>> RULE and some others to AND to fully satisfy 2)
>>
>> Is there any way? How can I do such things.
>>
>> Thanks in advance.
>>
>>
>>
>> On Tue, Dec 15, 2009 at 10:08 AM, alibaba <[log in to unmask]> wrote:
>>
>> Hello!
>>
>> First of all, thanks for reading.
>>
>> I need to add some Constraints to my GP program so each function that I
>> have can only accept a specific subset of children. For this, should specify
>> this on the params files or I have to add some checks on the
>> checkConstraints method for each GPNode.
>>
>> Does anybody know an example on the web about this that I can follow?
>>
>> Thanks in advance.
>>
>>
>>
>>
|