def testEstimate(self):
#Lets set up a simple model based on normal dist
abcParams = ABCParameters()
epsilonArray = numpy.array([0.5, 0.2, 0.1])
posteriorSampleSize = 20
#Lets get an empirical estimate of Sprime
model = NormalModel(abcMetrics)
model.setMu(theta[0])
model.setSigma(theta[1])
Sprime = abcMetrics.summary(model.simulate())
logging.debug(("Real summary statistic: " + str(Sprime)))
thetaDir = PathDefaults.getTempDir()
abcSMC = ABCSMC(epsilonArray, createNormalModel, abcParams, thetaDir)
abcSMC.maxRuns = 100000
abcSMC.setPosteriorSampleSize(posteriorSampleSize)
thetasArray = abcSMC.run()
thetasArray = numpy.array(thetasArray)
meanTheta = numpy.mean(thetasArray, 0)
logging.debug((thetasArray.shape))
logging.debug(thetasArray)
logging.debug(meanTheta)
print(thetasArray.shape[0], posteriorSampleSize)
#Note only mean needs to be similar
self.assertTrue(thetasArray.shape[0] >= posteriorSampleSize)
self.assertEquals(thetasArray.shape[1], 2)
self.assertTrue(numpy.linalg.norm(theta[0] - meanTheta[0]) < 0.2)
def testEstimate(self):
#Lets set up a simple model based on normal dist
abcParams = ABCParameters()
epsilonArray = numpy.array([0.5, 0.2, 0.1])
posteriorSampleSize = 20
#Lets get an empirical estimate of Sprime
model = NormalModel(abcMetrics)
model.setMu(theta[0])
model.setSigma(theta[1])
Sprime = abcMetrics.summary(model.simulate())
logging.debug(("Real summary statistic: " + str(Sprime)))
thetaDir = PathDefaults.getTempDir()
abcSMC = ABCSMC(epsilonArray, createNormalModel, abcParams, thetaDir)
abcSMC.maxRuns = 100000
abcSMC.setPosteriorSampleSize(posteriorSampleSize)
thetasArray = abcSMC.run()
thetasArray = numpy.array(thetasArray)
meanTheta = numpy.mean(thetasArray, 0)
logging.debug((thetasArray.shape))
logging.debug(thetasArray)
logging.debug(meanTheta)
print(thetasArray.shape[0], posteriorSampleSize)
#Note only mean needs to be similar
self.assertTrue(thetasArray.shape[0] >= posteriorSampleSize)
self.assertEquals(thetasArray.shape[1], 2)
self.assertTrue(numpy.linalg.norm(theta[0] - meanTheta[0]) < 0.2)
def testLoadParams(self):
try:
lmbda = 0.01
alterRegressor = PrimalRidgeRegression(lmbda)
egoRegressor = PrimalRidgeRegression(lmbda)
predictor = EgoEdgeLabelPredictor(alterRegressor, egoRegressor)
params = [0.1, 0.2]
paramFuncs = [egoRegressor.setLambda, alterRegressor.setLambda]
fileName = PathDefaults.getTempDir() + "tempParams.pkl"
predictor.saveParams(params, paramFuncs, fileName)
params2 = predictor.loadParams(fileName)
self.assertTrue(params2[0][0] == "apgl.predictors.PrimalRidgeRegression")
self.assertTrue(params2[0][1] == "setLambda")
self.assertTrue(params2[0][2] == 0.1)
self.assertTrue(params2[1][0] == "apgl.predictors.PrimalRidgeRegression")
self.assertTrue(params2[1][1] == "setLambda")
self.assertTrue(params2[1][2] == 0.2)
except IOError as e:
logging.warn(e)
def testLoadParams(self):
try:
lmbda = 0.01
alterRegressor = PrimalRidgeRegression(lmbda)
egoRegressor = PrimalRidgeRegression(lmbda)
predictor = EgoEdgeLabelPredictor(alterRegressor, egoRegressor)
params = [0.1, 0.2]
paramFuncs = [egoRegressor.setLambda, alterRegressor.setLambda]
fileName = PathDefaults.getTempDir() + "tempParams.pkl"
predictor.saveParams(params, paramFuncs, fileName)
params2 = predictor.loadParams(fileName)
self.assertTrue(
params2[0][0] == "apgl.predictors.PrimalRidgeRegression")
self.assertTrue(params2[0][1] == "setLambda")
self.assertTrue(params2[0][2] == 0.1)
self.assertTrue(
params2[1][0] == "apgl.predictors.PrimalRidgeRegression")
self.assertTrue(params2[1][1] == "setLambda")
self.assertTrue(params2[1][2] == 0.2)
except IOError as e:
logging.warn(e)
def testSaveParams(self):
try:
lmbda = 0.01
alterRegressor = PrimalRidgeRegression(lmbda)
egoRegressor = PrimalRidgeRegression(lmbda)
predictor = EgoEdgeLabelPredictor(alterRegressor, egoRegressor)
params = [0.1, 0.2]
paramFuncs = [egoRegressor.setLambda, alterRegressor.setLambda]
fileName = PathDefaults.getTempDir() + "tempParams.pkl"
predictor.saveParams(params, paramFuncs, fileName)
except IOError as e:
logging.warn(e)
def testSaveParams(self):
try:
lmbda = 0.01
alterRegressor = PrimalRidgeRegression(lmbda)
egoRegressor = PrimalRidgeRegression(lmbda)
predictor = EgoEdgeLabelPredictor(alterRegressor, egoRegressor)
params = [0.1, 0.2]
paramFuncs = [egoRegressor.setLambda, alterRegressor.setLambda]
fileName = PathDefaults.getTempDir() + "tempParams.pkl"
predictor.saveParams(params, paramFuncs, fileName)
except IOError as e:
logging.warn(e)
def profile(command, globalVars, localVars, numStats=30):
"""
Just profile the given command with the global and local variables
and print out the cumulative and function times.
"""
try:
import pstats
import cProfile
except ImportError:
raise ImportError("profile() requires pstats and cProfile")
tempDirectory = PathDefaults.getTempDir()
profileFileName = tempDirectory + "profile.cprof"
logging.info("Starting to profile ...")
cProfile.runctx(command, globalVars, localVars, profileFileName)
logging.info("Done")
stats = pstats.Stats(profileFileName)
stats.strip_dirs().sort_stats("cumulative").print_stats(numStats)
stats.strip_dirs().sort_stats("time").print_stats(numStats)
