elif p < 0.9:
nHidden = threeLayers()
else:
nHidden = fourLayers()
print 'Layers: ', nHidden
pretrainingEpochs=epochs()
pretrainingLearningRate=learningrate()
print 'Pretraining epochs: ', pretrainingEpochs
print 'Pretraining learning rate: ', pretrainingLearningRate
learningRate=learningrate()
numberEpochs=epochs()
print 'Fine tuning epochs: ', numberEpochs
print 'Fine tuning learning rate: ', learningRate
return HyperparametersDBN(learningRate=learningRate,
numberEpochs=numberEpochs,
pretrainingEpochs=pretrainingEpochs,
pretrainingLearningRate=pretrainingLearningRate,
nHidden=nHidden)
if __name__ == '__main__':
d = dataset.Mnist()
for i in range(1, 25):
print "Experiment %d" % i
logs = DBNClassifier.test_DBN(d, hyper())
print logs
print DBNClassifier.interpretObjectives(logs)
print ""
def hyper():
p = random.random()
nHidden = None
if p < 0.5:
nHidden = twoLayers()
elif p < 0.9:
nHidden = threeLayers()
else:
nHidden = fourLayers()
print 'Layers: ', nHidden
pretrainingEpochs=epochs()
pretrainingLearningRate=learningrate()
print 'Pretraining epochs: ', pretrainingEpochs
print 'Pretraining learning rate: ', pretrainingLearningRate
learningRate=learningrate()
numberEpochs=epochs()
print 'Fine tuning epochs: ', numberEpochs
print 'Fine tuning learning rate: ', learningRate
return HyperparametersDBN(learningRate=learningRate,
numberEpochs=numberEpochs,
pretrainingEpochs=pretrainingEpochs,
pretrainingLearningRate=pretrainingLearningRate,
nHidden=nHidden)
if __name__ == '__main__':
DBNClassifier.test_DBN(dataset.Mnist(), hyper())
def getObjectives(self):
objectives = DBNClassifier.test_DBN(self.data, self.hyper())
return DBNClassifier.interpretObjectives(objectives)
