def testLearnModel2(self):
#We want to make sure the learnt tree with gamma = 0 maximise the
#empirical risk
minSplit = 20
maxDepth = 3
gamma = 0.01
learner = PenaltyDecisionTree(minSplit=minSplit,
maxDepth=maxDepth,
gamma=gamma,
pruning=False)
#Vary sampleSize
numpy.random.seed(21)
learner.setSampleSize(1)
learner.learnModel(self.X, self.y)
error1 = learner.treeObjective(self.X, self.y)
numpy.random.seed(21)
learner.setSampleSize(5)
learner.learnModel(self.X, self.y)
error2 = learner.treeObjective(self.X, self.y)
numpy.random.seed(21)
learner.setSampleSize(10)
learner.learnModel(self.X, self.y)
error3 = learner.treeObjective(self.X, self.y)
self.assertTrue(error1 >= error2)
self.assertTrue(error2 >= error3)
#Now vary max depth
learner.gamma = 0
numpy.random.seed(21)
learner.setSampleSize(1)
learner.minSplit = 1
learner.maxDepth = 3
learner.learnModel(self.X, self.y)
predY = learner.predict(self.X)
error1 = Evaluator.binaryError(self.y, predY)
numpy.random.seed(21)
learner.maxDepth = 5
learner.learnModel(self.X, self.y)
predY = learner.predict(self.X)
error2 = Evaluator.binaryError(self.y, predY)
numpy.random.seed(21)
learner.maxDepth = 10
learner.learnModel(self.X, self.y)
predY = learner.predict(self.X)
error3 = Evaluator.binaryError(self.y, predY)
self.assertTrue(error1 >= error2)
self.assertTrue(error2 >= error3)
def testPrune(self):
startId = (0, )
minSplit = 20
maxDepth = 5
gamma = 0.05
learner = PenaltyDecisionTree(minSplit=minSplit,
maxDepth=maxDepth,
gamma=gamma,
pruning=False)
trainX = self.X[100:, :]
trainY = self.y[100:]
testX = self.X[0:100, :]
testY = self.y[0:100]
argsortX = numpy.zeros(trainX.shape, numpy.int)
for i in range(trainX.shape[1]):
argsortX[:, i] = numpy.argsort(trainX[:, i])
argsortX[:, i] = numpy.argsort(argsortX[:, i])
learner.tree = DictTree()
rootNode = DecisionNode(numpy.arange(trainX.shape[0]),
Util.mode(trainY))
learner.tree.setVertex(startId, rootNode)
learner.growTree(trainX, trainY, argsortX, startId)
learner.shapeX = trainX.shape
learner.predict(trainX, trainY)
learner.computeAlphas()
obj1 = learner.treeObjective(trainX, trainY)
size1 = learner.tree.getNumVertices()
#Now we'll prune
learner.prune(trainX, trainY)
obj2 = learner.treeObjective(trainX, trainY)
size2 = learner.tree.getNumVertices()
self.assertTrue(obj1 >= obj2)
self.assertTrue(size1 >= size2)
#Check there are no nodes with alpha>alphaThreshold
for vertexId in learner.tree.getAllVertexIds():
self.assertTrue(
learner.tree.getVertex(vertexId).alpha <=
learner.alphaThreshold)
def testLearnModel(self):
minSplit = 20
maxDepth = 3
gamma = 0.00
X, y = self.X, self.y
testX = X[100:, :]
testY = y[100:]
X = X[0:100, :]
y = y[0:100]
learner = PenaltyDecisionTree(minSplit=minSplit,
maxDepth=maxDepth,
gamma=gamma)
learner.learnModel(X, y)
tree = learner.getTree()
#Work out penalty cost
predY = learner.predict(X)
predTestY = learner.predict(testX)
n = float(X.shape[0])
d = X.shape[1]
T = tree.getNumVertices()
error = (1 - gamma) * numpy.sum(predY != y) / n
testError = numpy.sum(predTestY != testY) / float(testY.shape[0])
error += gamma * numpy.sqrt(T)
self.assertEquals(error, learner.treeObjective(X, y))
#Check if the values in the tree nodes are correct
for vertexId in tree.getAllVertexIds():
vertex = tree.getVertex(vertexId)
self.assertTrue(vertex.getValue() == 1.0
or vertex.getValue() == -1.0)
if tree.isNonLeaf(vertexId):
self.assertTrue(0 <= vertex.getFeatureInd() <= X.shape[1])
self.assertTrue(0 <= vertex.getError() <= 1)
def testGrowTree(self):
startId = (0, )
minSplit = 20
maxDepth = 3
gamma = 0.01
learner = PenaltyDecisionTree(minSplit=minSplit,
maxDepth=maxDepth,
gamma=gamma,
pruning=False)
trainX = self.X[100:, :]
trainY = self.y[100:]
testX = self.X[0:100, :]
testY = self.y[0:100]
argsortX = numpy.zeros(trainX.shape, numpy.int)
for i in range(trainX.shape[1]):
argsortX[:, i] = numpy.argsort(trainX[:, i])
argsortX[:, i] = numpy.argsort(argsortX[:, i])
learner.tree = DictTree()
rootNode = DecisionNode(numpy.arange(trainX.shape[0]),
Util.mode(trainY))
learner.tree.setVertex(startId, rootNode)
#Note that this matches with the case where we create a new tree each time
numpy.random.seed(21)
bestError = float("inf")
for i in range(20):
learner.tree.pruneVertex(startId)
learner.growTree(trainX, trainY, argsortX, startId)
predTestY = learner.predict(testX)
error = Evaluator.binaryError(predTestY, testY)
#print(Evaluator.binaryError(predTestY, testY), learner.tree.getNumVertices())
if error < bestError:
bestError = error
bestTree = learner.tree.copy()
self.assertTrue(learner.tree.depth() <= maxDepth)
for vertexId in learner.tree.nonLeaves():
self.assertTrue(
learner.tree.getVertex(vertexId).getTrainInds().shape[0] >=
minSplit)
bestError1 = bestError
learner.tree = bestTree
#Now we test growing a tree from a non-root vertex
numpy.random.seed(21)
for i in range(20):
learner.tree.pruneVertex((0, 1))
learner.growTree(trainX, trainY, argsortX, (0, 1))
self.assertTrue(
learner.tree.getVertex((0, )) == bestTree.getVertex((0, )))
self.assertTrue(
learner.tree.getVertex((0, 0)) == bestTree.getVertex((0, 0)))
predTestY = learner.predict(testX)
error = Evaluator.binaryError(predTestY, testY)
if error < bestError:
bestError = error
bestTree = learner.tree.copy()
#print(Evaluator.binaryError(predTestY, testY), learner.tree.getNumVertices())
self.assertTrue(bestError1 >= bestError)
def testComputeAlphas(self):
minSplit = 20
maxDepth = 3
gamma = 0.1
X, y = self.X, self.y
testX = X[100:, :]
testY = y[100:]
X = X[0:100, :]
y = y[0:100]
learner = PenaltyDecisionTree(minSplit=minSplit,
maxDepth=maxDepth,
gamma=gamma,
pruning=False)
learner.learnModel(X, y)
tree = learner.getTree()
rootId = (0, )
learner.tree.getVertex(rootId).setTestInds(numpy.arange(X.shape[0]))
learner.predict(X, y)
learner.computeAlphas()
#See if the alpha values of the nodes are correct
for vertexId in tree.getAllVertexIds():
subtreeLeaves = tree.leaves(vertexId)
subtreeError = 0
for subtreeLeaf in subtreeLeaves:
subtreeError += (
1 - gamma) * tree.getVertex(subtreeLeaf).getTestError()
n = float(X.shape[0])
d = X.shape[1]
T = tree.getNumVertices()
subtreeError /= n
subtreeError += gamma * numpy.sqrt(T)
T2 = T - len(tree.subtreeIds(vertexId)) + 1
vertexError = (1 -
gamma) * tree.getVertex(vertexId).getTestError() / n
vertexError += gamma * numpy.sqrt(T2)
self.assertAlmostEquals((subtreeError - vertexError),
tree.getVertex(vertexId).alpha)
if tree.isLeaf(vertexId):
self.assertEquals(tree.getVertex(vertexId).alpha, 0.0)
#Let's check the alpha of the root node via another method
rootId = (0, )
T = 1
(n, d) = X.shape
n = float(n)
vertexError = (1 - gamma) * numpy.sum(y != Util.mode(y)) / n
pen = gamma * numpy.sqrt(T)
vertexError += pen
T = tree.getNumVertices()
treeError = (1 - gamma) * numpy.sum(y != learner.predict(X)) / n
pen = gamma * numpy.sqrt(T)
treeError += pen
alpha = treeError - vertexError
self.assertAlmostEqual(alpha, tree.getVertex(rootId).alpha)
