// test generating random individuals
func TestGenerate(t *testing.T) {
pset := initPset(false)
pset.Add(V(0), V(1))
gen := gp.GenRamped(pset, 1, 3)
gp.SetSeed(0)
for i := 0; i < 10; i++ {
ind := gen.Generate()
res := ind.Code.Eval(V(6), V(7))
t.Log(ind.Code, ind.Code.Format(), "(6,7) =>", res)
}
}
// main GP routine
func main() {
// get options
var maxSize, maxDepth int
var dataFile string
flag.IntVar(&maxSize, "size", 0, "maximum tree size - zero for none")
flag.IntVar(&maxDepth, "depth", 0, "maximum tree depth - zero for none")
flag.StringVar(&dataFile, "trainset", "poly.dat", "file with training function")
opts := util.DefaultOptions
util.ParseFlags(&opts)
// create primitive set
ercMin, ercMax, trainSet := getData(dataFile)
pset := gp.CreatePrimSet(1, "x")
pset.Add(num.Add, num.Sub, num.Mul, num.Div)
pset.Add(num.Ephemeral("ERC", ercGen(ercMin, ercMax)))
// setup model
problem := &gp.Model{
PrimitiveSet: pset,
Generator: gp.GenRamped(pset, 1, 3),
PopSize: opts.PopSize,
Fitness: fitnessFunc(trainSet),
Offspring: gp.Tournament(opts.TournSize),
Mutate: gp.MutUniform(gp.GenGrow(pset, 0, 2)),
MutateProb: opts.MutateProb,
Crossover: gp.CxOnePoint(),
CrossoverProb: opts.CrossoverProb,
Threads: opts.Threads,
}
if maxDepth > 0 {
problem.AddDecorator(gp.DepthLimit(maxDepth))
}
if maxSize > 0 {
problem.AddDecorator(gp.SizeLimit(maxSize))
}
problem.PrintParams("== GP Symbolic Regression for ", dataFile, "==")
// run
logger := stats.NewLogger(opts.MaxGen, opts.TargetFitness)
if opts.Plot {
gp.GraphDPI = "60"
logger.RegisterPlot("graph", plotTarget(trainSet), plotBest(trainSet))
stats.MainLoop(problem, logger, ":8080", "../web")
} else {
fmt.Println()
logger.PrintStats = true
logger.PrintBest = opts.Verbose
problem.Run(logger)
}
}