def test_boost(self):
listAttr = randlist(0, 5, 10)
listInst = [dtree.Instance(listAttr, True) for _ in xrange(100)]
listInstFalse = random.sample(listInst, 10)
for inst in listInstFalse:
inst.fLabel = False
listInstCopy = [inst.copy() for inst in listInst]
br = dtree.boost(listInstCopy)
dblWeightExpected = dtree.classifier_weight(0.1)
self.assertAlmostEqual(br.listDblCferWeight[0], dblWeightExpected)
def test_boost(self):
listAttr = randlist(0,5,10)
listInst = [dtree.Instance(listAttr, True) for _ in xrange(100)]
listInstFalse = random.sample(listInst,10)
for inst in listInstFalse:
inst.fLabel = False
listInstCopy = [inst.copy() for inst in listInst]
br = dtree.boost(listInstCopy)
dblWeightExpected = dtree.classifier_weight(0.1)
self.assertAlmostEqual(br.listDblCferWeight[0], dblWeightExpected)
def test_one_round_boost(self):
fxnGen = build_consistent_generator()
cInst = 100
listInst = fxnGen(cInst)
for inst in listInst:
inst.listAttrs[0] = int(inst.fLabel)
listInstIncorrect = random.sample(listInst, cInst / 10)
for inst in listInstIncorrect:
inst.fLabel = not inst.listAttrs[0]
inst.dblWeight = 0.1
dt, dblError, dblCferWeight = dtree.one_round_boost(listInst, 1)
self.assertTrue(is_stump(dt))
self.assertAlmostEqual(1.0 / 91.0, dblError)
self.assertAlmostEqual(dtree.classifier_weight(dblError),
dblCferWeight)
self.assertAlmostEqual(1.0, sum([inst.dblWeight for inst in listInst]))
def test_one_round_boost(self):
fxnGen = build_consistent_generator()
cInst = 100
listInst = fxnGen(cInst)
for inst in listInst:
inst.listAttrs[0] = int(inst.fLabel)
listInstIncorrect = random.sample(listInst,cInst/10)
for inst in listInstIncorrect:
inst.fLabel = not inst.listAttrs[0]
inst.dblWeight = 0.1
dt,dblError,dblCferWeight = dtree.one_round_boost(listInst,1)
self.assertTrue(is_stump(dt))
self.assertAlmostEqual(1.0/91.0, dblError)
self.assertAlmostEqual(dtree.classifier_weight(dblError),
dblCferWeight)
self.assertAlmostEqual(1.0, sum([inst.dblWeight for inst in listInst]))
def test_classifier_weight(self):
dblError = random.random()
dblWeight = dtree.classifier_weight(dblError)
dblFrac = math.exp(2.0 * dblWeight)
self.assertAlmostEqual(dblError, 1.0 / (dblFrac + 1.0))
def test_classifier_weight(self):
dblError = random.random()
dblWeight = dtree.classifier_weight(dblError)
dblFrac = math.exp(2.0*dblWeight)
self.assertAlmostEqual(dblError, 1.0/(dblFrac + 1.0))
