def setUp(self):
self.baseDir = os.path.join(RDConfig.RDCodeDir, 'ML', 'test_data')
self.dbName = RDConfig.RDTestDatabase
self.details = BuildComposite.SetDefaults()
self.details.dbName = self.dbName
self.details.dbUser = RDConfig.defaultDBUser
self.details.dbPassword = RDConfig.defaultDBPassword
def test1(self):
""" basics """
self.details.tableName = 'ferro_quant'
refComposName = 'ferromag_quant_10.pkl'
refCompos = pickle.load(open(os.path.join(self.baseDir,refComposName),
'rb'))
# first make sure the data are intact
self._init(refCompos)
compos = BuildComposite.RunIt(self.details,saveIt=0)
self.compare(compos,refCompos)
def test7(self): """ Test composite of naive bayes""" self.details.tableName = 'ferro_noquant' refComposName = 'ferromag_NaiveBayes.pkl' pklFile = open(os.path.join(self.baseDir,refComposName), 'rb') refCompos = pickle.load(pklFile) self._init(refCompos,copyBounds=1) self.details.useTrees = 0 self.details.useNaiveBayes = 1 self.details.mEstimateVal = 20.0 self.details.qBounds = [0] + [2]*6 + [0] compos = BuildComposite.RunIt(self.details, saveIt= 0) self.compare(compos,refCompos)
def test6(self):
""" auto bounds with a real valued activity"""
self.details.tableName = 'ferro_noquant_realact'
refComposName = 'ferromag_auto_10_3.pkl'
refCompos = pickle.load(open(os.path.join(self.baseDir,refComposName),
'rb'))
# first make sure the data are intact
self._init(refCompos,copyBounds=1)
self.details.limitDepth = 3
self.details.nModels = 10
self.details.activityBounds=[0.5]
compos = BuildComposite.RunIt(self.details,saveIt=0)
self.compare(compos,refCompos)
def test4(self):
""" more trees """
self.details.tableName = 'ferro_quant'
refComposName = 'ferromag_quant_50_3.pkl'
refCompos = pickle.load(open(os.path.join(self.baseDir,refComposName),
'rb'))
# first make sure the data are intact
self._init(refCompos)
self.details.limitDepth = 3
self.details.nModels = 50
compos = BuildComposite.RunIt(self.details,saveIt=0)
self.compare(compos,refCompos)
def test3(self):
""" depth limit + less greedy """
self.details.tableName = 'ferro_quant'
refComposName = 'ferromag_quant_10_3_lessgreedy.pkl'
refCompos = pickle.load(open(os.path.join(self.baseDir,refComposName),
'rb'))
# first make sure the data are intact
self._init(refCompos)
self.details.limitDepth = 3
self.details.lessGreedy = 1
compos = BuildComposite.RunIt(self.details,saveIt=0)
self.compare(compos,refCompos)
def test7(self):
""" Test composite of naive bayes"""
self.details.tableName = 'ferro_noquant'
refComposName = 'ferromag_NaiveBayes.pkl'
with open(os.path.join(self.baseDir, refComposName), 'r') as pklTFile:
buf = pklTFile.read().replace('\r\n', '\n').encode('utf-8')
pklTFile.close()
with io.BytesIO(buf) as pklFile:
refCompos = pickle.load(pklFile)
self._init(refCompos, copyBounds=1)
self.details.useTrees = 0
self.details.useNaiveBayes = 1
self.details.mEstimateVal = 20.0
self.details.qBounds = [0] + [2] * 6 + [0]
compos = BuildComposite.RunIt(self.details, saveIt=0)
self.compare(compos, refCompos)
def test2(self):
""" depth limit """
self.details.tableName = 'ferro_quant'
refComposName = 'ferromag_quant_10_3.pkl'
with open(os.path.join(self.baseDir, refComposName), 'r') as pklTF:
buf = pklTF.read().replace('\r\n', '\n').encode('utf-8')
pklTF.close()
with io.BytesIO(buf) as pklF:
refCompos = pickle.load(pklF)
# first make sure the data are intact
self._init(refCompos)
self.details.limitDepth = 3
compos = BuildComposite.RunIt(self.details, saveIt=0)
self.compare(compos, refCompos)
def test6(self):
""" auto bounds with a real valued activity"""
self.details.tableName = 'ferro_noquant_realact'
refComposName = 'ferromag_auto_10_3.pkl'
with open(os.path.join(self.baseDir, refComposName), 'r') as pklTF:
buf = pklTF.read().replace('\r\n', '\n').encode('utf-8')
pklTF.close()
with io.BytesIO(buf) as pklF:
refCompos = pickle.load(pklF)
# first make sure the data are intact
self._init(refCompos, copyBounds=1)
self.details.limitDepth = 3
self.details.nModels = 10
self.details.activityBounds = [0.5]
compos = BuildComposite.RunIt(self.details, saveIt=0)
self.compare(compos, refCompos)
def test1_basics(self):
# """ basics """
self.details.tableName = 'ferro_quant'
refComposName = 'ferromag_quant_10.pkl'
with open(os.path.join(self.baseDir, refComposName), 'r') as pklTF:
buf = pklTF.read().replace('\r\n', '\n').encode('utf-8')
pklTF.close()
with io.BytesIO(buf) as pklF:
refCompos = pickle.load(pklF)
# first make sure the data are intact
self._init(refCompos)
compos = BuildComposite.RunIt(self.details, saveIt=0)
# pickle.dump(compos,open(os.path.join(self.baseDir,refComposName), 'wb'))
# with open(os.path.join(self.baseDir,refComposName), 'rb') as pklF:
# refCompos = pickle.load(pklF)
self.compare(compos, refCompos)
def GrowIt(details,composite,progressCallback=None,
saveIt=1,setDescNames=0,data=None):
""" does the actual work of building a composite model
**Arguments**
- details: a _CompositeRun.CompositeRun_ object containing details
(options, parameters, etc.) about the run
- composite: the composite model to grow
- progressCallback: (optional) a function which is called with a single
argument (the number of models built so far) after each model is built.
- saveIt: (optional) if this is nonzero, the resulting model will be pickled
and dumped to the filename specified in _details.outName_
- setDescNames: (optional) if nonzero, the composite's _SetInputOrder()_ method
will be called using the results of the data set's _GetVarNames()_ method;
it is assumed that the details object has a _descNames attribute which
is passed to the composites _SetDescriptorNames()_ method. Otherwise
(the default), _SetDescriptorNames()_ gets the results of _GetVarNames()_.
- data: (optional) the data set to be used. If this is not provided, the
data set described in details will be used.
**Returns**
the enlarged composite model
"""
details.rundate = time.asctime()
if data is None:
fName = details.tableName.strip()
if details.outName == '':
details.outName = fName + '.pkl'
if details.dbName == '':
data = DataUtils.BuildQuantDataSet(fName)
elif details.qBounds != []:
details.tableName = fName
data = details.GetDataSet()
else:
data = DataUtils.DBToQuantData(details.dbName,fName,quantName=details.qTableName,
user=details.dbUser,password=details.dbPassword)
nExamples = data.GetNPts()
seed = composite._randomSeed
DataUtils.InitRandomNumbers(seed)
testExamples = []
if details.shuffleActivities == 1:
DataUtils.RandomizeActivities(data,shuffle=1,runDetails=details)
elif details.randomActivities == 1:
DataUtils.RandomizeActivities(data,shuffle=0,runDetails=details)
namedExamples = data.GetNamedData()
trainExamples = namedExamples
nExamples = len(trainExamples)
message('Training with %d examples'%(nExamples))
message('\t%d descriptors'%(len(trainExamples[0])-2))
nVars = data.GetNVars()
nPossibleVals = composite.nPossibleVals
attrs = range(1,nVars+1)
if details.useTrees:
from rdkit.ML.DecTree import CrossValidate,PruneTree
if details.qBounds != []:
from rdkit.ML.DecTree import BuildQuantTree
builder = BuildQuantTree.QuantTreeBoot
else:
from rdkit.ML.DecTree import ID3
builder = ID3.ID3Boot
driver = CrossValidate.CrossValidationDriver
pruner = PruneTree.PruneTree
if setDescNames:
composite.SetInputOrder(data.GetVarNames())
composite.Grow(trainExamples,attrs,[0]+nPossibleVals,
buildDriver=driver,
pruner=pruner,
nTries=details.nModels,pruneIt=details.pruneIt,
lessGreedy=details.lessGreedy,needsQuantization=0,
treeBuilder=builder,nQuantBounds=details.qBounds,
startAt=details.startAt,
maxDepth=details.limitDepth,
progressCallback=progressCallback,
silent=not _verbose)
else:
from rdkit.ML.Neural import CrossValidate
driver = CrossValidate.CrossValidationDriver
composite.Grow(trainExamples,attrs,[0]+nPossibleVals,nTries=details.nModels,
buildDriver=driver,needsQuantization=0)
composite.AverageErrors()
composite.SortModels()
modelList,counts,avgErrs = composite.GetAllData()
counts = numpy.array(counts)
avgErrs = numpy.array(avgErrs)
composite._varNames = data.GetVarNames()
for i in range(len(modelList)):
modelList[i].NameModel(composite._varNames)
# do final statistics
weightedErrs = counts*avgErrs
averageErr = sum(weightedErrs)/sum(counts)
devs = (avgErrs - averageErr)
devs = devs * counts
devs = numpy.sqrt(devs*devs)
avgDev = sum(devs)/sum(counts)
if _verbose:
message('# Overall Average Error: %%% 5.2f, Average Deviation: %%% 6.2f'%(100.*averageErr,100.*avgDev))
if details.bayesModel:
composite.Train(trainExamples,verbose=0)
badExamples = []
if not details.detailedRes:
if _verbose:
message('Testing all examples')
wrong = BuildComposite.testall(composite,namedExamples,badExamples)
if _verbose:
message('%d examples (%% %5.2f) were misclassified'%(len(wrong),100.*float(len(wrong))/float(len(namedExamples))))
_runDetails.overall_error = float(len(wrong))/len(namedExamples)
if details.detailedRes:
if _verbose:
message('\nEntire data set:')
resTup = ScreenComposite.ShowVoteResults(range(data.GetNPts()),data,composite,
nPossibleVals[-1],details.threshold)
nGood,nBad,nSkip,avgGood,avgBad,avgSkip,voteTab = resTup
nPts = len(namedExamples)
nClass = nGood+nBad
_runDetails.overall_error = float(nBad) / nClass
_runDetails.overall_correct_conf = avgGood
_runDetails.overall_incorrect_conf = avgBad
_runDetails.overall_result_matrix = repr(voteTab)
nRej = nClass-nPts
if nRej > 0:
_runDetails.overall_fraction_dropped = float(nRej)/nPts
return composite