def processResults(experiment):
import os
import shutil
import numpy as np
import pbsGridWalker.tools.plotutils as tplt
tfs.makeDirCarefully('results', maxBackups=100)
def fitnessFileName(gp):
return 'NI' + str(gp['newIndividualsPerGeneration']) + '_IP' + gp['initialPopulationType'] + '_fitness'
def columnExtractor(gp):
outFile = fitnessFileName(gp)
subprocess.call('cut -d \' \' -f 2 bestIndividual*.log | tail -n +4 | tr \'\n\' \' \' >> ../results/' + outFile, shell=True)
subprocess.call('echo >> ../results/' + outFile, shell=True)
experiment.executeAtEveryGridPointDir(columnExtractor)
os.chdir('results')
xlabel = 'Generations'
ylimit = None
yscale = 'lin'
xscale = 'lin'
margins = 0.5
xlimit = None
alpha = 0.3
title = None
def robustLoadTxt(ffn):
arr0 = []
with open(ffn, 'r') as ff:
for line in ff:
vals = map(float, line.split())
arr0.append(vals)
minlen = min([ len(vs) for vs in arr0 ])
arr1 = []
for vs in arr0:
arr1.append(vs[:minlen])
arr1 = np.array(arr1)
return arr1
# dataDict = { str(mg): np.loadtxt(fitnessFileName(mg)) for mg in [0.1, 0.3, 0.5, 0.7, 0.9] }
# dataDict = { str(mg): robustLoadTxt(fitnessFileName(mg)) for mg in [0.1, 0.3, 0.5, 0.7, 0.9] }
# tplt.plotAverageTimeSeries(dataDict, 'Error', 'mg.png',
# title=title, legendLocation=None, xlabel=xlabel,
# xlimit=xlimit, ylimit=ylimit, figsize=(2.5,4), xscale=xscale, yscale=yscale, margins=margins)
for ip in ['sparse', 'random']:
dataDict = { ('NI='+str(ni)): robustLoadTxt(fitnessFileName({'initialPopulationType': ip, 'newIndividualsPerGeneration': ni})) for ni in [1,4,16] }
tplt.plotAllTimeSeries(dataDict, 'Fitness', 'tu_IP' + ip + '.png',
title=title, legendLocation=1, xlabel=xlabel,
xlimit=xlimit, ylimit=ylimit, xscale=xscale, yscale=yscale, margins=margins, alpha=alpha)
# def plotAllTSForInitalPopulationType(initPopType):
# title = None
# dataDict = {attachments[x]: -1.*np.loadtxt(fitnessFileName(x, initPopType)) for x in attachments.keys()}
# tplt.plotAverageTimeSeries(dataDict, 'Error', 'errorComparison_GENS50_IP' + initPopType + '.png', title=title, legendLocation=None, xlabel=xlabel, xlimit=50, ylimit=ylimit, figsize=(2.5,4), xscale=xscale, yscale=yscale, margins=margins)
# tplt.plotAllTimeSeries(dataDict, 'Error', 'errorAllTrajectories_GEN50_IP' + initPopType + '.png', title=title, legendLocation=None, xlabel=xlabel, xlimit=50, ylimit=ylimit, figsize=(2.5,4), xscale=xscale, yscale=yscale, margins=margins, alpha=alpha)
# tplt.plotAverageTimeSeries(dataDict, 'Error', 'errorComparison_IP' + initPopType + '.png', title=title, legendLocation=1, xlabel=xlabel, xlimit=500, ylimit=ylimit, xscale=xscale, yscale=yscale, margins=margins)
# tplt.plotAllTimeSeries(dataDict, 'Error', 'errorAllTrajectories_IP' + initPopType + '.png', title=title, legendLocation=1, xlabel=xlabel, xlimit=500, ylimit=ylimit, xscale=xscale, yscale=yscale, margins=margins, alpha=alpha)
# for ip in initialPopulationTypes:
# plotAllTSForNew(ip)
os.chdir('..')
def processResults(experiment):
import os
import numpy as np
import pbsGridWalker.tools.plotutils as tplt
tfs.makeDirCarefully('results', maxBackups=100)
# We'll take a look at some parameters vs relative mutation rate at several stages (generation counts) along the evolutionary process
# Linear stages
stagesToConsider = 5
stages = tal.splitIntegerRangeIntoStages(0, evsAdditionalParams['genStopAfter'], stagesToConsider)
##### Extracting and plotting the distance to the maximally modular morphology (MMM) for various values relative mutation rate #####
# mmmmdist and similar abbreviations stand for "minimal distance to the maximally modular morphology" (across the Pareto front)
xlabel = r'$P_{mm}$'
fieldNames = [ 'gen {}'.format(st) for st in stages ]
def generateMinMMMDistTimeSlices(gridPoint):
return [ gctools.minParetoFrontHammingDistanceToMMM(gen) for gen in stages ]
tplt.plotComputationVariableAgainstParameter(experiment, 'mmmmd', generateMinMMMDistTimeSlices, 'probabilityOfMutatingClass0',
fieldNames=fieldNames, xlabel=xlabel, ylabel=r'$\mu$')
def generateFitnessTimeSlices(gridPoint):
bestIndividualData = np.loadtxt('bestIndividual{}.log'.format(gridPoint['randomSeed']))
fitnessData = []
for genRec in range(bestIndividualData.shape[0]):
gen = int(bestIndividualData[genRec,0])
if gen in stages:
fitnessData.append(bestIndividualData[genRec,1]) # WILL break if the best individual records are not in the order of increasing generation
return fitnessData
tplt.plotComputationVariableAgainstParameter(experiment, 'error', generateFitnessTimeSlices, 'probabilityOfMutatingClass0',
fieldNames=fieldNames, transform=lambda x: x, yscale='lin', xlabel=xlabel, ylabel=r'$\log_10 E$', strips='conf95', forcedYLabelPos=[0.05,1])
def processResults(experiment):
import os
import shutil
import numpy as np
import pbsGridWalker.tools.plotutils as tplt
tfs.makeDirCarefully('results', maxBackups=100)
def fitnessFileName(sensAttType, initPopType):
return 'SA' + sensAttType + '_IP' + initPopType + '_fitness'
def columnExtractor(gp):
outFile = fitnessFileName(gp['sensorAttachmentType'], gp['initialPopulationType'])
subprocess.call('cut -d \' \' -f 2 bestIndividual*.log | tail -n +4 | tr \'\n\' \' \' >> ../results/' + outFile, shell=True)
subprocess.call('echo >> ../results/' + outFile, shell=True)
experiment.executeAtEveryGridPointDir(columnExtractor)
os.chdir('results')
xlabel = 'Generations'
ylimit = None
yscale = 'log'
xscale = 'log'
margins = 0.5
xlimit = 500
alpha=0.3
def plotAllTSForInitalPopulationType(initPopType):
title = None # 'Fitness time series for the two types of sensors attachment'
dataDict = {attachments[x]: -1.*np.loadtxt(fitnessFileName(x, initPopType)) for x in attachments.keys()}
tplt.plotAverageTimeSeries(dataDict, 'Error', 'errorComparison_GENS50_IP' + initPopType + '.png', title=title, legendLocation=None, xlabel=xlabel, xlimit=50, ylimit=ylimit, figsize=(2.5,4), xscale=xscale, yscale=yscale, margins=margins)
tplt.plotAllTimeSeries(dataDict, 'Error', 'errorAllTrajectories_GEN50_IP' + initPopType + '.png', title=title, legendLocation=None, xlabel=xlabel, xlimit=50, ylimit=ylimit, figsize=(2.5,4), xscale=xscale, yscale=yscale, margins=margins, alpha=alpha)
tplt.plotAverageTimeSeries(dataDict, 'Error', 'errorComparison_IP' + initPopType + '.png', title=title, legendLocation=1, xlabel=xlabel, xlimit=500, ylimit=ylimit, xscale=xscale, yscale=yscale, margins=margins)
tplt.plotAllTimeSeries(dataDict, 'Error', 'errorAllTrajectories_IP' + initPopType + '.png', title=title, legendLocation=1, xlabel=xlabel, xlimit=500, ylimit=ylimit, xscale=xscale, yscale=yscale, margins=margins, alpha=alpha)
for ip in initialPopulationTypes:
plotAllTSForInitalPopulationType(ip)
os.chdir('..')
def processResults(experiment):
tfs.makeDirCarefully('results', maxBackups=100)
plotTTCvsMMMD(experiment)
def gridFileNamePrefix(gridPoint):
if gridPoint['probabilityOfMutatingClass0'] == 0.2:
if gridPoint['compositeClass0'] == 'integerVectorSymmetricRangeMutations':
return 'move_sensor_by_1_segment'
elif gridPoint['compositeClass0'] == 'integerVectorRandomJumps':
return 're-generate_morphology'
elif gridPoint['probabilityOfMutatingClass0'] == 0 and gridPoint['compositeClass0'] == 'integerVectorSymmetricRangeMutations':
return 'no_morphological_mutation'
raise ValueError('Wrong point {} in the non-RS grid'.format(gridPoint))
plotErrorTSs(experiment, gridFileNamePrefix)
plotMinMMMDistTSs(experiment, gridFileNamePrefix)
def processResults(experiment):
import os
import shutil
import numpy as np
import pbsGridWalker.tools.plotutils as tplt
tfs.makeDirCarefully('results', maxBackups=100)
def fitnessFileName(paramsDict):
gid = 'NOGID'
if paramsDict['lineageInjectionPeriod'] == 30000:
if paramsDict['individual'] == 'ctrnnDiscreteWeightsFleetOfIdenticalsFixedFitness':
gid = '1'
elif paramsDict['individual'] == 'ctrnnDiscreteWeightsFleetOfIdenticalsEvolvableFitness':
gid = '2'
elif paramsDict['lineageInjectionPeriod'] == 50:
if paramsDict['mutatedLineagesRatio'] == 0.:
if paramsDict['evolver'] == 'ageFunction':
if paramsDict['initialPopulationType'] == 'random':
gid = '3'
elif paramsDict['initialPopulationType'] == 'sparse':
gid = '6'
elif paramsDict['evolver'] == 'ageFunctionSparsityBiased':
if paramsDict['initialPopulationType'] == 'random':
gid = '4'
elif paramsDict['initialPopulationType'] == 'sparse':
gid = '5'
elif paramsDict['mutatedLineagesRatio'] == 1.0:
if paramsDict['lineageMutationType'] == 'individualClassDefault':
gid = '7'
elif paramsDict['lineageMutationType'] == 'randomJump':
gid = '8'
if gid == 'NOGID':
raise ValueError('Unclassified parameter dictionary {}'.format(paramsDict))
return 'gid{}'.format(gid)
def columnExtractor(gp):
print('Extracting fitness times series to {}...'.format(fitnessFileName(gp)))
outFile = fitnessFileName(gp)
subprocess.call('cut -d \' \' -f 2 bestIndividual*.log | tail -n +4 | tr \'\n\' \' \' >> ../results/' + outFile, shell=True)
subprocess.call('echo >> ../results/' + outFile, shell=True)
experiment.executeAtEveryGridPointDir(columnExtractor)
def plotAll():
os.chdir('results')
xlabel = 'Generations'
ylimit = None
yscale = 'lin'
xscale = 'lin'
margins = 0.5
xlimit = 6000
alpha = 0.3
title = None
# figsize = (2.5,4)
figsize = None
striptype = 'conf95'
filenames = [ 'gid{}'.format(x) for x in range(1,9) ]
dataDict = { fn: np.loadtxt(fn) for fn in filenames }
comp3 = { tsn: dataDict[fn] for tsn,fn in {'no turning': 'gid3', 'smooth turning': 'gid7', 'abrupt turning': 'gid8'}.items() }
comparisons = { 'comp3': comp3 }
for e,d in comparisons.items():
tplt.plotAverageTimeSeries(d, 'Fitness', e + '-avg.png',
title=title, legendLocation=4, xlabel=xlabel,
xlimit=xlimit, ylimit=ylimit, figsize=figsize, xscale=xscale, yscale=yscale, margins=margins, strips=striptype)
tplt.plotAllTimeSeries(d, 'Fitness', e + '-all.png', alpha=alpha,
title=title, legendLocation=4, xlabel=xlabel,
xlimit=xlimit, ylimit=ylimit, figsize=figsize, xscale=xscale, yscale=yscale, margins=margins)
os.chdir('..')
plotAll()
