I’m Neio

Noted by Neio

I’m do some experiments with the method of Genetic Programming (GP). Open BEAGLE is the first choice that my advicer recommended to me to do the experiments because of it’s efficiency and some other quality. However, the references about Open BEAGLE are limited on internet. In this case, I would noted something useful here that was not mentioned on it’s official Wiki.

Here are the parameters to run the simulation (”HaRec” is the simulation program I have written):

usage: HaRec [-OBparameter=value] …
HaRec [-OBparam1=value1,param2=value2, ... ,paramN=valueN] …

Supported parameters:

ec.conf.dump <String> (def: “”)
Filename used to dump the configuration. A configuration dump means that a configuration file is written with the evolver (including the composing operators) and the register (including the registered parameters and their default values). No evolution is conducted on a configuration dump.
An empty string means no dump.

ec.conf.file <String> (def: “”)
The name of a configuration file containing evolver and parameter values.
A typical configuration file can be created with parameter “ec.conf.dump”.

ec.hof.demesize <UInt> (def: 0)
Number of individuals kept in each deme’s hall-of-fame (best individuals so far). Note that a hall-of-fame contains only copies of the best individuals so far and is not used by the evolution process.

ec.hof.vivasize <UInt> (def: 1)
Number of individuals kept in vivarium’s hall-of-fame (best individuals so far). Note that a hall-of-fame contains only copies of the best individuals so far and is not used by the evolution process.

ec.init.seedsfile <String> (def: “”)
Name of file to use for seeding the evolution with crafted individual. An empty string means no seeding.

ec.mig.interval <UInt> (def: 1)
Interval between each migration, in number of generations. An interval of 0 disables migration.

ec.mig.size <UInt> (def: 5)
Number of individuals migrating between each deme, at a each migration.

ec.pop.size <UIntArray> (def: 100)
Number of demes and size of each deme of the population. The format of an UIntArray is S1,S2,…,Sn, where Si is the ith value. The size of the UIntArray is the number of demes present in the vivarium, while each value of the vector is the size of the corresponding deme.

ec.rand.seed <ULong> (def: 0)
Randomizer seed. A zero value means that the seed will be initialized using the current system time.

ec.repro.prob <Float> (def: 0.1)
Probability that an individual is reproducted as is, without modification. This parameter is useful only in selection and initialization operators that are composing a breeder tree.

ec.sel.tournsize <UInt> (def: 2)
Number of participants for tournament selection.

ec.term.maxgen <UInt> (def: 50)
Maximum number of generations for the evolution.

gp.cx.distrpb <Float> (def: 0.9)
Probability that a crossover point is a branch (node with sub-trees).
Value of 1.0 means that all crossover points are branches, and value of 0.0 means that all crossover points are leaves.

gp.cx.indpb <Float> (def: 0.9)
Individual crossover probability at each generation.

gp.init.maxargs <UIntArray> (def: 0/2)
Maximum number of arguments in GP tree. Tree arguments are is usually useful with ADFs (and similar stuff).

gp.init.maxdepth <UInt> (def: 5)
Maximum depth for newly initialized trees.

gp.init.maxtree <UInt> (def: 1)
Maximum number of GP tree in newly initialized individuals. More than one tree is usually useful with ADFs (and other ADx).

gp.init.minargs <UIntArray> (def: 0/2)
Minimum number of arguments in GP tree. Tree arguments are is usually useful with ADFs a(nd similar stuff).

gp.init.mindepth <UInt> (def: 2)
Minimum depth for newly initialized trees.

gp.init.mintree <UInt> (def: 1)
Minimum number of GP tree in newly initialized individuals. More than one tree is usually useful with ADFs (and other ADx).

gp.mutshrink.indpb <Float> (def: 0.05)
Shrink mutation probability for an individual. Shrink mutation consists in replacing a branch (a node with one or more arguments) with one of his child node. This erases the chosen node and the other child nodes.

gp.mutsst.distrpb <Float> (def: 0.5)
Probability that a swap subtree is internal (the mutation occurs between three points, where the 2nd point is in the 1st point’s subtree, and the 3rd point is in the 2nd point’s subtree) vs being external (the mutation occurs between two points, where both points are not within the other’s
subtree). Value of 1.0 means that the swap subtrees mutations are all internal while value of 0.0 means that swap subtrees mutations are all external.

gp.mutsst.indpb <Float> (def: 0.0)
Swap subtree mutation probability for an individual. A swap subtree mutation consists to swap two subtrees of a tree in an individual.

gp.mutstd.indpb <Float> (def: 0.05)
Standard mutation probability for an individual. A standard mutation replaces a sub-tree with a randomly generated one.

gp.mutstd.maxdepth <UInt> (def: 5)
Maximum depth for standard mutation. A standard mutation replaces a sub-tree with a randomly generated one.

gp.mutswap.distrpb <Float> (def: 0.5)
Probability that a swap mutation point is a branch (node with sub-trees).
Value of 1.0 means that all swap mutation points are branches, and value of 0.0 means that all swap mutation points are leaves. Swap mutation consists in exchanging the primitive associated to a node by one having the same number of arguments.

gp.mutswap.indpb <Float> (def: 0.05)
Swap mutation probability for an individual. Swap mutation consists in exchanging the primitive associated to a node by one having the same number of arguments.

gp.tree.maxdepth <UInt> (def: 17)
Maximum allowed depth for the trees.

gp.try <UInt> (def: 2)
Maximum number of attempts to modify a GP tree in a genetic operation. As there is topological constraints on GP trees (i.e. tree depth limit), it is often necessary to try a genetic operation several times.

help
Shows the Open BEAGLE specific command-line usage and detailed parameter
descriptions.

lg.console.level <UInt> (def: 2)
Log level used for console output generation. Log levels available are:
(0) no log, (1) basic logs, (2) stats, (3) general informations, (4) details on operations, (5) trace of the algorithms, (6) verbose, (7) debug (enabled only in full debug mode).

lg.file.level <UInt> (def: 3)
Log level used for file output generation. Log levels available are: (0) no log, (1) basic logs, (2) stats, (3) general informations, (4) details on operations, (5) trace of the algorithms, (6) verbose, (7) debug (enabled only in full debug mode).

lg.file.name <String> (def: “beagle.log”)
Filename in which messages are outputed. An empty string file name means no output is done to a file.

lg.show.class <Bool> (def: 0)
Flag whether class name is outputed in the logs.

lg.show.level <Bool> (def: 0)
Flag whether logging level in outputed in the logs.

lg.show.type <Bool> (def: 0)
Flag whether message type is outputed in the logs.

ms.restart.file <String> (def: “”)
Name of the milestone file from which the evolution should be restarted.
An empty string means no restart.

ms.write.interval <UInt> (def: 0)
Milestone saving interval (in number of generations). When zero, only the last generation milestone is saved.

ms.write.over <Bool> (def: 1)
If true, this flag indicates that old milestones should be over-written.
Otherwise, each milestone has a different suffix.

ms.write.perdeme <Bool> (def: 0)
If true, this flag indicates that separate milestones should be written after each demes processing. Otherwise milestones are written after the processing of a complete populations.

ms.write.prefix <String> (def: “beagle”)
Prefix used to name the evolution milestone files. An empty string means no milestone.

usage
Shows the Open BEAGLE specific command-line usage.