def analyze(ctx):
ctxobj = ctx.obj
ga = GenAlg(
chromoSize=3 * num_sprayers,
dtype=np.float32,
range=dataRanges,
fitnessFcn=calcFitness,
crossoverFcn=crossover,
userDataI=cutout_region, # user-provided cut-out region
userDataF=scoring, # user-provided scoring system
size=128,
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
minOrMax='max',
showBest=0)
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
ga.describe()
# Run it !!
ga.evolve(1)
ga.cudaAnalysis()
def main():
random.seed()
ga = GenAlg( size=popsize,
elitism = 0.03,
crossover = 0.47,
pureMutation = 0.50,
selectionFcn = GenAlgOps.simpleSelectionParentPct,
crossoverFcn = GenAlgOps.crossover22,
mutationFcn = GenAlgOps.mutateFew,
# for pure-mutation of all chromos .. no need to run tournament selection
# pureMutationSelectionFcn = lambda x: [0,0],
# pureMutationFcn = GenAlgOps.mutateAll,
chromoClass = MyChromo,
minOrMax = 'max',
showBest = 0,
# optional params ..
params = {
'mutateNum': 3, # for mutateFew .. make 2 mutations each time
'parentPct': 0.50 # for parent-pct .. only top 50% of chromos are eligible as parents
},
)
ga.describe()
#print( 'random state', random.getstate() )
#
# if a data-file exists, we load it
if( os.path.isfile('ga_mathsq.dat') ):
pop = IoOps.loadPopulation( ga, 'ga_mathsq.dat' )
ga.appendToPopulation( pop )
print( 'Read init data from file ('+str(len(pop))+' chromos)')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print( 'Created random init data' )
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range( outeriter ):
ga.evolve( inneriter )
sys.stdout.write( '.' )
sys.stdout.flush()
#
# all done ... output final results
print( "\nfinal best chromo: " + str(ga.population[0].fitness) )
ga.population[0].showGrid()
debuglevel = 100
ga.population[0].calcFitness()
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
IoOps.savePopulation( ga, 'ga_mathsq.dat' )
print('Final data stored to file (rm ga_mathsq.dat to start fresh)')
def main():
ga = GenAlg(
size=100,
elitism=0.10,
crossover=0.60,
pureMutation=0.30,
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
#crossoverFcn = GenAlgOps.crossover22,
#mutationFcn = GenAlgOps.mutateFew,
#pureMutationSelectionFcn = GenAlgOps.simpleSelection,
#pureMutationFcn = GenAlgOps.mutateFew,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
minOrMax='max',
showBest=0,
)
ga.describe()
#
# if a data-file exists, we load it
if (False and os.path.isfile('ga_hello.dat')):
pop = IoOps.loadPopulation(ga, 'ga_hello.dat')
ga.appendToPopulation(pop)
print('Read init data from file (' + str(len(pop)) + ' chromos)')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(10):
ga.evolve(10)
# give some running feedback on our progress
print('iter ' + str(i) + ", best chromo:")
for i in range(10):
print(ga.population[i])
#
# all done ... output final results
print("\nfinal best chromos:")
for i in range(10):
print(ga.population[i])
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
IoOps.savePopulation(ga, 'ga_hello.dat')
print('Final data stored to file (rm ga_hello.dat to start fresh)')
def test1(ctx):
ctxobj = ctx.obj
verbose = ctxobj.get('verbose')
ga = GenAlg(
size=4,
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
crossoverFcn=MyCrossover212,
mutationFcn=GenAlgOps.mutateFew,
# for pure-mutation of all chromos .. no need to run tournament selection
#pureMutationSelectionFcn = lambda x: [0,0],
#pureMutationFcn = GenAlgOps.mutateAll,
pureMutationSelectionFcn=GenAlgOps.simpleSelection2,
pureMutationFcn=MyMutate,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
minOrMax='min',
showBest=0,
# optional params ..
params={
'mutateNum': ctxobj.get('mutatenum'),
'parentPct': 0.50,
})
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
# simulate a cross-over
mother = ga.population[0]
print('mother', str(mother))
father = ga.population[1]
print('father', str(father))
children = ga.crossoverFcn(mother, father, ga.params)
for child in children:
print('child', str(child))
asm_prog = [
'ROM0', 'XYZ0', 'MPY', 'ROM1', 'XYZ1', 'MPY', 'ADD', 'ROM2', 'ADD'
]
prog = ga.population[2].cpu.compile(asm_prog)
print('prog', prog)
prog.extend([1, 1, 1]) # add the 3 rom/coeffs
# but make these the last N steps of prog, so the output _is_ the prog output
ga.population[2].data[-len(prog):] = prog
fit = ga.population[2].calcFitness()
print('code', ga.population[2].cpu.show_prog(show_pc=False, nl='/'))
print('code', ga.population[2].cpu.show_prog_as_func())
print('fitness', fit)
def test1(ctx):
ctxobj = ctx.obj
verbose = ctxobj.get('verbose')
ga = GenAlg(
size=4,
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
crossoverFcn=GenAlgOps.crossover12,
mutationFcn=GenAlgOps.mutateNone,
# for pure-mutation of all chromos .. no need to run tournament selection
#pureMutationSelectionFcn = lambda x: [0,0],
#pureMutationFcn = GenAlgOps.mutateAll,
pureMutationSelectionFcn=GenAlgOps.simpleSelectionParentPct,
pureMutationFcn=GenAlgOps.mutateAll,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
minOrMax='min',
showBest=0,
# optional params ..
params={
'mutateNum': ctxobj.get('mutatenum'),
'parentPct': 0.50,
})
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
mother = ga.population[0]
mother.fitness = mother.calcFitness()
print('mother', str(mother))
#print( ' ', mother.dataRange )
print(' ', mother.cpu.show_prog_as_func())
father = ga.population[1]
father.fitness = father.calcFitness()
print('father', str(father))
#print( ' ', father.dataRange )
print(' ', father.cpu.show_prog_as_func())
# simulate a cross-over
children = ga.crossoverFcn(mother, father, ga.params)
for child in children:
child.fitness = child.calcFitness()
print('child', str(child))
def main():
global debuglevel
ga = GenAlg(size=250,
elitismPct=0.10,
crossoverPct=0.30,
mutationPct=0.60,
parentsPct=0.50,
chromoClass=MyChromo,
minOrMax='max',
showBest=0)
#
# if a data-file exists, we load it
if (os.path.isfile('ga_maze.dat')):
ga.loadPopulation('ga_maze.dat')
print('Read init data from file')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(10):
ga.evolve(100)
sys.stdout.write('.')
sys.stdout.flush()
#
# all done ... output final results
print("\nfinal best chromo: " + str(ga.population[0].fitness))
ga.population[0].showGrid()
debuglevel = 100
ga.population[0].calcFitness()
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
ga.savePopulation('ga_maze.dat')
print('Final data stored to file (rm ga_maze.dat to start fresh)')
def main():
sims = [('max', MyChromoPos), ('min', MyChromoPos), ('max', MyChromoNeg),
('min', MyChromoNeg)]
for s in sims:
print(s)
ga = GenAlg(
size=10,
elitism=0.10,
crossover=0.60,
pureMutation=0.30,
parentsPct=0.80,
chromoClass=s[1],
#selectionFcn = GenAlgOps.tournamentSelection,
#crossoverFcn = GenAlgOps.crossover22,
#mutationFcn = GenAlgOps.mutateFew,
#pureMutationSelectionFcn = GenAlgOps.simpleSelection,
#pureMutationFcn = GenAlgOps.mutateFew,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
minOrMax=s[0],
showBest=0,
)
ga.initPopulation()
for i in range(10):
if (s[1] == MyChromoNeg):
ga.population[i].data[0] = -i - 1
else:
ga.population[i].data[0] = i + 1
ga.calcFitness()
ga.sortPopulation()
printGaPopulation(ga)
counts = [0 for i in range(10)]
for i in range(10000):
idx1, idx2 = GenAlgOps.rouletteWheelSelection(ga)
counts[idx1] = counts[idx1] + 1
counts[idx2] = counts[idx2] + 1
print(' roulette counts = ' + str(counts))
counts = [0 for i in range(10)]
for i in range(10000):
idx1, idx2 = GenAlgOps.rankSelection(ga)
counts[idx1] = counts[idx1] + 1
counts[idx2] = counts[idx2] + 1
print(' rank counts = ' + str(counts))
def main():
ga = GenAlg(chromoSize=4,
dtype=np.int32,
range=(0, 10),
fitnessFcn=calcFitness,
crossoverFcn=crossover2,
size=1024,
elitism=0.10,
crossover=0.60,
pureMutation=0.30,
minOrMax='max',
showBest=0)
#
# if a data-file exists, we load it
if (os.path.isfile('ga_coins.dat')):
pop = IoOps.loadPopulation(ga, 'ga_coins.dat')
ga.loadPopulation(pop)
print('Read init data from file (' + str(len(pop)) + ' chromos)')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(10):
ga.evolve(10)
# give some running feedback on our progress
print('iter ' + str(i) + ", best chromo:")
print(ga.population[range(4)], ga.fitnessVals[0])
#
# all done ... output final results
print("\nfinal best chromos:")
print(ga.population[range(4)], ga.fitnessVals[0])
print(ga.population[range(4, 8)], ga.fitnessVals[1])
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
IoOps.savePopulation(ga, 'ga_coins.dat')
print('Final data stored to file (rm ga_coins.dat to start fresh)')
def main():
ga = GenAlg(
size=10,
elitism=0.10,
crossover=0.60,
pureMutation=0.30,
parentsPct=0.80,
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
#crossoverFcn = GenAlgOps.crossover22,
#mutationFcn = GenAlgOps.mutateFew,
#pureMutationSelectionFcn = GenAlgOps.simpleSelection,
#pureMutationFcn = GenAlgOps.mutateFew,
feasibleSolnFcn=GenAlgOps.disallowDupes,
minOrMax='max',
showBest=0,
)
ga.initPopulation()
ga.calcFitness()
child = MyChromo()
child.data[0] = ga.population[4].data[0]
child.fitness = ga.population[4].fitness
feas = GenAlgOps.disallowDupes(ga, child)
print('unsorted population; data present; feas=', feas)
child.data[0] = ga.population[4].data[0] + 9
child.fitness = ga.population[4].fitness + 9
feas = GenAlgOps.disallowDupes(ga, child)
print('unsorted population; data not present; feas=', feas)
ga.sortPopulation()
child.data[0] = ga.population[4].data[0]
child.fitness = ga.population[4].fitness
feas = GenAlgOps.disallowDupes(ga, child)
print('sorted population; data at pos=4; feas=', feas)
child.data[0] = ga.population[4].data[0] + 9
child.fitness = ga.population[4].fitness + 9
feas = GenAlgOps.disallowDupes(ga, child)
print('sorted population; data not present; feas=', feas)
def run(self):
tid = self.tid
num_pes = self.num_pes
parMgr = self.parMgr
# TODO: maybe this is an MPI_INIT kind of functionality?
# : e.g. store this worker-tid into parMgr; init per-worker values
parMgr.tid = tid
print('TID' + str(tid) + ' started (out of ' + str(num_pes) + ')')
sys.stdout.flush()
# TODO: calculate the splitting across the PEs
# otherwise, init the gen-alg library from scratch
ga = GenAlg(size=100,
elitismPct=0.10,
crossoverPct=0.30,
mutationPct=0.50,
migrationPct=0.10,
migrationFcn=self.migrationFcn,
chromoClass=MyChromo,
minOrMax='max',
showBest=0)
ga.initPopulation()
#ga.loadPopulation( 'ga_data.dat' )
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(10):
print('pop_sz=' + str(len(ga.population)))
ga.evolve(10)
# give some running feedback on our progress
print('iter ' + str(i) + ", best chromo:")
for i in range(1):
print(ga.population[i])
# at this point, each PE's population is sorted
rtn = parMgr.collect(ga.population[:10])
bestVals = [x for y in rtn for x in y]
bestVals.sort()
#
# all done ... output final results
print("\nfinal best chromos:")
for i in range(10):
print(bestVals[i])
def sweep(ctx, sweep_iters, append, outfile):
ctxobj = ctx.obj
verbose = ctxobj.get('verbose')
epoch_it = ctxobj.get('epoch_it')
inner_it = ctxobj.get('inner_it')
ga = GenAlg(
chromoSize=3 * num_sprayers,
dtype=np.float32,
range=dataRanges,
fitnessFcn=calcFitness,
crossoverFcn=crossover,
userDataI=cutout_region, # user-provided cut-out region
userDataF=scoring, # user-provided scoring system
size=ctxobj.get('popsize'),
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
parentPct=ctxobj.get('parent_pct'),
minOrMax='max',
showBest=0)
if (append):
fp_out = open(append, 'a')
elif (outfile):
fp_out = open(outfile, 'w')
else:
fp_out = sys.stdout
if (verbose > 0):
print('Epoch/Inner iters:', epoch_it, inner_it)
print('Sweep iters:', sweep_iters)
for m in range(sweep_iters):
ga.initPopulation()
with click.progressbar(range(epoch_it)) as bar:
for i in bar:
ga.evolve(inner_it)
fp_out.write('%d, %d, %d, %f\n' %
(ga.population_sz, epoch_it, inner_it, ga.fitnessVals[0]))
fp_out.close()
def main():
ga = GenAlg(
size=100,
elitism=0.10,
crossover=0.60,
pureMutation=0.30,
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
#crossoverFcn = GenAlgOps.crossover22,
#mutationFcn = GenAlgOps.mutateFew,
#pureMutationSelectionFcn = GenAlgOps.simpleSelection,
#pureMutationFcn = GenAlgOps.mutateFew,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
minOrMax='min',
showBest=0,
)
# init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(100):
ga.evolve(10)
# give some running feedback on our progress
#print( 'iter '+str(i) + ", best chromo:" )
#for i in range(10):
# print( ga.population[i] )
#
# all done ... output final results
print("\nfinal best chromos:")
for i in range(3):
print(ga.population[i])
def run(self):
print('TID start')
sys.stdout.flush()
tid = self.tid
num_pes = self.num_pes
commMgr = self.commMgr
# TODO: maybe this is an MPI_INIT kind of functionality?
print('TID' + str(tid) + ' started (out of ' + str(num_pes) + ')')
sys.stdout.flush()
# TODO: calculate the splitting across the PEs
# otherwise, init the gen-alg library from scratch
ga = GenAlg(size=1000,
elitism=0.10,
crossover=0.50,
pureMutation=0.35,
migration=0.05,
migrationSendFcn=self.migrationSendFcn,
migrationRecvFcn=self.migrationRecvFcn,
parents=0.80,
chromoClass=MyChromo,
minOrMax='max',
showBest=0)
#
# if a pickle-file exists, we load it
if (os.path.isfile('ga_sprinkler.dat')):
# in a parallel sim, we need to load different chunks of the data-file
# into each TID .. so we need to manage this through the commMgr
# pop = IoOps.loadPopulation( ga, 'ga_sprinkler.dat' )
pop = commMgr.loadPopulation(ga, 'ga_sprinkler.dat')
ga.appendToPopulation(pop)
print('Read init data from file (' + str(len(pop)) + ' chromos)')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(10):
ga.evolve(10)
# give some running feedback on our progress
print('iter ' + str(i) + ", best chromo:")
for i in range(1):
print(ga.population[i])
# at this point, each PE's population is sorted
rtn = commMgr.collect(ga.population[:10])
# rtn is a list of lists ... flatten it
bestVals = [x for y in rtn for x in y]
bestVals = ga.sortPopulationList(bestVals)
#
# all done ... output final results
if (tid == 0):
print("\nfinal best chromos:")
for i in range(1):
print(bestVals[i])
bestVals[0].drawImage()
commMgr.savePopulation(ga, 'ga_sprinkler.dat')
def thresh(ctx, sweep_iters, threshold, append, outfile):
ctxobj = ctx.obj
verbose = ctxobj.get('verbose')
epoch_it = ctxobj.get('epoch_it')
inner_it = ctxobj.get('inner_it')
if (append):
fp_out = open(append, 'a')
elif (outfile):
fp_out = open(outfile, 'w')
else:
fp_out = sys.stdout
maxspray = 0.0
for i in range(num_sprayers):
maxspray = maxspray + 3.1415 * dataRanges[3 * i + 2][1] * dataRanges[
3 * i + 2][1]
maxfield = 100 * 100 - cutout_region[0] * cutout_region[1]
mx = max(scoring)
if (threshold < 1):
threshold = threshold * mx * maxspray
# find all the pairs of vals to test
test_list = []
for m in range(sweep_iters):
for pp in range(1, 11):
for ps in [1024, 2048, 4096, 8192, 16384, 32768]:
test_list.append((ps, pp / 10, m))
if (verbose > 0):
print('Epoch/Inner iters:', epoch_it, inner_it)
print('Sweep iters:', sweep_iters)
print('Threshold: %.2f' % (threshold))
print('Max spray: %.2f ( %d )' % (mx * maxspray, mx * maxfield))
print('Total num tests:', len(test_list))
# run the tests w/ progress bar
with click.progressbar(test_list) as bar:
for params in bar:
# print( 'params', params )
ga = GenAlg(
chromoSize=3 * num_sprayers,
dtype=np.float32,
range=dataRanges,
fitnessFcn=calcFitness,
crossoverFcn=crossover,
userDataI=cutout_region, # user-provided cut-out region
userDataF=scoring, # user-provided scoring system
size=params[0],
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
parentPct=params[1],
minOrMax='max',
showBest=0)
ga.initPopulation()
totalits = 0
flag = 0
for e in range(epoch_it):
ga.evolve(inner_it)
totalits = totalits + inner_it
if (ga.fitnessVals[0] >= threshold):
flag = 1
break
fp_out.write(
'%d, %d, %d, %d, %.2f, %d, %d, %.2f\n' %
(ga.population_sz, ga.elitism, ga.crossover, ga.pureMutation,
ga.parent_pct, flag, totalits, ga.fitnessVals[0]))
fp_out.flush()
fp_out.close()
def run(self):
tid = self.tid
num_pes = self.num_pes
parMgr = self.parMgr
# TODO: maybe this is an MPI_INIT kind of functionality?
# : e.g. store this worker-tid into parMgr; init per-worker values
parMgr.tid = tid
print('TID' + str(tid) + ' started (out of ' + str(num_pes) + ')')
sys.stdout.flush()
# TODO: calculate the splitting across the PEs
ga = GenAlg(
size=250,
elitismPct=0.10,
crossoverPct=0.30,
mutationPct=0.50,
migrationPct=0.10,
#selectionFcn = GenAlgOps.tournamentSelection,
#crossoverFcn = GenAlgOps.crossover22,
#mutationFcn = GenAlgOps.mutateFew,
#pureMutationSelectionFcn = GenAlgOps.simpleSelection,
#pureMutationFcn = GenAlgOps.mutateFew,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
migrationSendFcn=self.migrationSend,
migrationRecvFcn=self.migrationRecv,
chromoClass=MyChromo,
minOrMax='min',
showBest=0)
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
t0 = time.time()
for i in range(15):
ga.evolve(10)
# give some running feedback on our progress
#print( str(i) + " best chromo:" )
#for i in range(1):
# print( ga.population[i] )
t1 = time.time()
print('TID' + str(tid) + ' time = ' + str(t1 - t0) + ' sec')
# at this point, each PE's population is sorted
rtn = parMgr.collect(ga.population[:10])
bestVals = [x for y in rtn for x in y]
bestVals.sort()
#
# all done ... output final results
if (tid == 0):
print("\nfinal best chromos:")
for i in range(1):
pop = bestVals[i].data
# print( ga.population[i] )
pop_ct = bestVals[i].population_per_district()
for j in range(len(pop_ct)):
ct = pop_ct[j]
x = pop[2 * j]
y = pop[2 * j + 1]
print(' %10.4f %10.4f : %10d' % (x, y, ct))
bestVals[0].drawImage()
def run(ctx, save_out, load_in):
ctxobj = ctx.obj
verbose = ctxobj.get('verbose')
epoch_it = ctxobj.get('epoch_it')
inner_it = ctxobj.get('inner_it')
ga = GenAlg(
chromoSize=3 * num_sprayers,
dtype=np.float32,
range=dataRanges,
fitnessFcn=calcFitness,
crossoverFcn=crossover,
userDataI=cutout_region, # user-provided cut-out region
userDataF=scoring, # user-provided scoring system
size=ctxobj.get('popsize'),
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
parentPct=ctxobj.get('parent_pct'),
minOrMax='max',
showBest=0)
#
# if a pickle-file exists, we load it
if (load_in):
pop = IoOps.loadPopulation(ga, 'ga_sprinkler.dat')
ga.loadPopulation(pop)
print('Read init data from file (' + str(len(pop)) + ' chromos)')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
if (verbose > 0):
ga.describe()
print('Epoch/Inner iters:', epoch_it, inner_it)
maxspray = 0.0
for i in range(num_sprayers):
maxspray = maxspray + 3.1415 * dataRanges[
3 * i + 2][1] * dataRanges[3 * i + 2][1]
maxfield = 100 * 100 - cutout_region[0] * cutout_region[1]
mx = max(scoring)
print('Max spray: %.2f ( %d )' % (mx * maxspray, mx * maxfield))
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
with click.progressbar(range(epoch_it)) as bar:
for i in bar:
ga.evolve(inner_it)
#
# all done ... output final results
print("\nfinal best chromos:")
print(ga.population[range(3 * num_sprayers)], ga.fitnessVals[0])
drawImage(ga.population[range(3 * num_sprayers)])
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
if (save_out):
IoOps.savePopulation(ga, 'ga_sprinkler.dat')
print('Final data stored to file (rm ga_sprinkler.dat to start fresh)')
def main():
ga = GenAlg(
size=100,
elitism=0.10,
crossover=0.60,
pureMutation=0.30,
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
#crossoverFcn = GenAlgOps.crossover22,
#mutationFcn = GenAlgOps.mutateFew,
#pureMutationSelectionFcn = GenAlgOps.simpleSelection,
#pureMutationFcn = GenAlgOps.mutateFew,
#feasibleSolnFcn = GenAlgOps.disallowDupes,
minOrMax='min',
showBest=0,
)
#
# if a pickle-file exists, we load it
if (os.path.isfile('ga_voting2.dat')):
pop = IoOps.loadPopulation(ga, 'ga_voting2.dat')
ga.appendToPopulation(pop)
print('Read init data from file (' + str(len(pop)) + ' chromos)')
if (len(pop) < ga.population_sz):
# we need to fill this out with random data
pop = IoOps.randomPopulation(ga, ga.population_sz - len(pop))
ga.appendToPopulation(pop)
print(' appended ' + str(len(pop)) + ' random chromos')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(10):
ga.evolve(10)
# give some running feedback on our progress
print(str(i) + " best chromo:")
for i in range(1):
print(ga.population[i])
#
# all done ... output final results
print("\nfinal best chromos:")
for i in range(1):
pop = ga.population[i]
# print( ga.population[i] )
pop_ct = pop.population_per_district()
for j in range(len(pop_ct)):
ct = pop_ct[j]
x = pop.data[2 * j]
y = pop.data[2 * j + 1]
print(' %10.4f %10.4f : %10d' % (x, y, ct))
ga.population[0].drawImage()
ga.population[0].drawImage(showCurrent=True)
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
IoOps.savePopulation(ga, 'ga_voting2.dat')
print('Final data stored to file (rm ga_voting2.dat to start fresh)')
def run(ctx):
ctxobj = ctx.obj
verbose = ctxobj.get('verbose')
epoch_it = ctxobj.get('epoch_it')
inner_it = ctxobj.get('inner_it')
ga = GenAlg(
size=ctxobj.get('popsize'),
elitism=ctxobj.get('elitism'),
crossover=ctxobj.get('crossover'),
pureMutation=ctxobj.get('puremutation'),
chromoClass=MyChromo,
#selectionFcn = GenAlgOps.tournamentSelection,
crossoverFcn=MyCrossover212,
mutationFcn=GenAlgOps.mutateNone,
# for pure-mutation of all chromos .. no need to run tournament selection
#pureMutationSelectionFcn = lambda x: [0,0],
#pureMutationFcn = GenAlgOps.mutateAll,
pureMutationSelectionFcn=GenAlgOps.simpleSelectionParentPct,
pureMutationFcn=MyMutate,
feasibleSolnFcn=GenAlgOps.disallowDupes,
minOrMax='min',
showBest=0,
# optional params ..
params={
'mutateNum': ctxobj.get('mutatenum'),
'parentPct': 0.50,
})
#
# if a data-file exists, we load it
if (ctxobj.get('load_in')):
pop = IoOps.loadPopulation(ga, 'ga_cpu.dat')
ga.appendToPopulation(pop)
print('Read init data from file (' + str(len(pop)) + ' chromos)')
else:
# otherwise, init the gen-alg library from scratch
ga.initPopulation()
print('Created random init data')
# add some 'good' candidate solns ... ROM0*XYZ0, ROM1*XYZ1, etc.
cpu = ga.population[0].cpu
prog = cpu.compile(['ROM0', 'XYZ0', 'MPY', 'ROM1', 'XYZ1', 'MPY'])
for i in range(0, prog_size - 6):
ga.population[i].data[-len(prog):] = prog
if (verbose > 0):
ga.describe()
print('Chromo size: %d :: %d %d' %
(len(ga.population[0].data), prog_size, rom_size))
print('Epoch/Inner iters:', epoch_it, inner_it)
print('Instruction set:',
' '.join(ga.population[0].cpu.PARSEops.keys()))
#
# Run it !!
# : we'll just do 10 epochs of 10 steps each
for i in range(epoch_it):
print('it=', i, time.time())
ga.evolve(inner_it)
# give some running feedback on our progress
txt = ''
for j in range(10):
txt = txt + ' %d' % (ga.population[j].fitness)
print('iter ' + str(i) + ", best fitnesses:" + txt)
print(' ' + ga.population[0].cpu.show_prog(show_pc=False, nl='/') +
' :: ' + ga.population[0].cpu.dump_rommemory())
print(' ' + ga.population[0].cpu.show_prog_as_func())
#
# all done ... output final results
print("\nfinal best chromos:")
for i in range(5):
#print( ga.population[i] )
print(' fit=%d' % (ga.population[i].fitness))
print(' ' + ga.population[i].cpu.show_prog(show_pc=False, nl='/') +
' :: ' + ga.population[i].cpu.dump_rommemory())
print(' ' + ga.population[i].cpu.show_prog_as_func())
#
# we'll always save the pickle-file, just delete it
# if you want to start over from scratch
if (ctxobj.get('save_out')):
IoOps.savePopulation(ga, 'ga_cpu.dat')
print('Final data stored to file (rm ga_cpu.dat to start fresh)')