def test_leave_pair_out(self):
#compares holdout and leave-pair-out
start = [0, 2, 3, 5]
end = [1, 3, 6, 8]
for X in [self.Xtrain1, self.Xtrain2]:
for Y in [self.Ytrain1, self.Ytrain2]:
#LPO with linear kernel
rls1 = GlobalRankRLS(X, Y, regparam = 7.0, bias=3.0)
lpo_start, lpo_end = rls1.leave_pair_out(start, end)
ho_start, ho_end = [], []
for i in range(len(start)):
P = rls1.holdout([start[i], end[i]])
ho_start.append(P[0])
ho_end.append(P[1])
ho_start = np.array(ho_start)
ho_end = np.array(ho_end)
assert_allclose(ho_start, lpo_start)
assert_allclose(ho_end, lpo_end)
#LPO Gaussian kernel
rls1 = GlobalRankRLS(X, Y, regparam = 11.0, kenerl="PolynomialKernel", coef0=1, degree=3)
lpo_start, lpo_end = rls1.leave_pair_out(start, end)
ho_start, ho_end = [], []
for i in range(len(start)):
P = rls1.holdout([start[i], end[i]])
ho_start.append(P[0])
ho_end.append(P[1])
ho_start = np.array(ho_start)
ho_end = np.array(ho_end)
assert_allclose(ho_start, lpo_start)
assert_allclose(ho_end, lpo_end)
def test_leave_pair_out(self):
#compares holdout and leave-pair-out
start = [0, 2, 3, 5]
end = [1, 3, 6, 8]
for X in [self.Xtrain1, self.Xtrain2]:
for Y in [self.Ytrain1, self.Ytrain2]:
#LPO with linear kernel
rls1 = GlobalRankRLS(X, Y, regparam=7.0, bias=3.0)
lpo_start, lpo_end = rls1.leave_pair_out(start, end)
ho_start, ho_end = [], []
for i in range(len(start)):
P = rls1.holdout([start[i], end[i]])
ho_start.append(P[0])
ho_end.append(P[1])
ho_start = np.array(ho_start)
ho_end = np.array(ho_end)
assert_allclose(ho_start, lpo_start)
assert_allclose(ho_end, lpo_end)
#LPO Gaussian kernel
rls1 = GlobalRankRLS(X,
Y,
regparam=11.0,
kenerl="PolynomialKernel",
coef0=1,
degree=3)
lpo_start, lpo_end = rls1.leave_pair_out(start, end)
ho_start, ho_end = [], []
for i in range(len(start)):
P = rls1.holdout([start[i], end[i]])
ho_start.append(P[0])
ho_end.append(P[1])
ho_start = np.array(ho_start)
ho_end = np.array(ho_end)
assert_allclose(ho_start, lpo_start)
assert_allclose(ho_end, lpo_end)
def test_holdout(self):
for X in [self.Xtrain1, self.Xtrain2]:
for Y in [self.Ytrain1, self.Ytrain2]:
m = X.shape[0]
hoindices = [3, 5, 8, 10, 17, 21]
hocompl = list(set(range(m)) - set(hoindices))
#Holdout with linear kernel
rls1 = GlobalRankRLS(X, Y)
rls2 = GlobalRankRLS(X[hocompl], Y[hocompl])
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Holdout with bias
rls1 = GlobalRankRLS(X, Y, bias=3.0)
rls2 = GlobalRankRLS(X[hocompl], Y[hocompl], bias=3.0)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Fast regularization
for i in range(-5, 5):
rls1.solve(2**i)
rls2.solve(2**i)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Kernel holdout
rls1 = GlobalRankRLS(X, Y, kernel="GaussianKernel", gamma=0.01)
rls2 = GlobalRankRLS(X[hocompl],
Y[hocompl],
kernel="GaussianKernel",
gamma=0.01)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
for i in range(-15, 15):
rls1.solve(2**i)
rls2.solve(2**i)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Incorrect indices
I = [0, 3, 100]
self.assertRaises(IndexError, rls1.holdout, I)
I = [-1, 0, 2]
self.assertRaises(IndexError, rls1.holdout, I)
I = [1, 1, 2]
self.assertRaises(IndexError, rls1.holdout, I)
def test_holdout(self):
for X in [self.Xtrain1, self.Xtrain2]:
for Y in [self.Ytrain1, self.Ytrain2]:
m = X.shape[0]
hoindices = [3, 5, 8, 10, 17, 21]
hocompl = list(set(range(m)) - set(hoindices))
#Holdout with linear kernel
rls1 = GlobalRankRLS(X, Y)
rls2 = GlobalRankRLS(X[hocompl], Y[hocompl])
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Holdout with bias
rls1 = GlobalRankRLS(X, Y, bias = 3.0)
rls2 = GlobalRankRLS(X[hocompl], Y[hocompl], bias = 3.0)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Fast regularization
for i in range(-5, 5):
rls1.solve(2**i)
rls2.solve(2**i)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Kernel holdout
rls1 = GlobalRankRLS(X, Y, kernel = "GaussianKernel", gamma = 0.01)
rls2 = GlobalRankRLS(X[hocompl], Y[hocompl], kernel = "GaussianKernel", gamma = 0.01)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
for i in range(-15, 15):
rls1.solve(2**i)
rls2.solve(2**i)
P1 = rls1.holdout(hoindices)
P2 = rls2.predict(X[hoindices])
assert_allclose(P1, P2)
#Incorrect indices
I = [0, 3, 100]
self.assertRaises(IndexError, rls1.holdout, I)
I = [-1, 0, 2]
self.assertRaises(IndexError, rls1.holdout, I)
I = [1,1,2]
self.assertRaises(IndexError, rls1.holdout, I)
def train_rls():
#Trains RLS with default parameters (regparam=1.0, kernel='LinearKernel')
X_train, Y_train, X_test, Y_test = load_housing()
#generate fold partition, arguments: train_size, k, random_seed
folds = random_folds(len(Y_train), 5, 10)
learner = GlobalRankRLS(X_train, Y_train)
perfs = []
for fold in folds:
P = learner.holdout(fold)
c = cindex(Y_train[fold], P)
perfs.append(c)
perf = np.mean(perfs)
print("5-fold cross-validation cindex %f" % perf)
P_test = learner.predict(X_test)
print("test cindex %f" % cindex(Y_test, P_test))
def train_rls():
#Trains RLS with default parameters (regparam=1.0, kernel='LinearKernel')
X_train, Y_train, X_test, Y_test = load_housing()
#generate fold partition, arguments: train_size, k, random_seed
folds = random_folds(len(Y_train), 5, 10)
learner = GlobalRankRLS(X_train, Y_train)
perfs = []
for fold in folds:
P = learner.holdout(fold)
c = cindex(Y_train[fold], P)
perfs.append(c)
perf = np.mean(perfs)
print("5-fold cross-validation cindex %f" %perf)
P_test = learner.predict(X_test)
print("test cindex %f" %cindex(Y_test, P_test))
def testAllPairsRankRLS(self):
print(
"Testing the cross-validation routines of the GlobalRankRLS module.\n\n"
)
np.random.seed(100)
floattype = np.float64
m, n, h = 30, 200, 10
Xtrain = np.random.rand(m, n)
trainlabels = np.random.rand(m, h)
trainkm = np.dot(Xtrain, Xtrain.T)
ylen = 1
L = np.mat(m * np.eye(m) - np.ones((m, m), dtype=floattype))
hoindices = [5, 7]
hoindices3 = [5, 7, 9]
hocompl = list(set(range(m)) - set(hoindices))
hocompl3 = list(set(range(m)) - set(hoindices3))
loglambdas = range(-5, 5)
for j in range(0, len(loglambdas)):
regparam = 2.**loglambdas[j]
print("\nRegparam 2^%1d" % loglambdas[j])
Kcv = trainkm[np.ix_(hocompl, hocompl)]
Ycv = trainlabels[hocompl]
Ktest = trainkm[np.ix_(hocompl, hoindices)]
Xcv = Xtrain[hocompl]
Xtest = Xtrain[hoindices]
Lcv = np.mat((m - 2) * np.eye(m - 2) -
np.ones((m - 2, m - 2), dtype=floattype))
oind = 1
rpool = {}
rpool['Y'] = Ycv
rpool['X'] = Xcv
rpool['regparam'] = regparam
naivedualrls = GlobalRankRLS(**rpool)
naivedualrls.solve(regparam)
hopreds = []
hopred = naivedualrls.predictor.predict(Xtest)
print(str(hopred[0, oind]) + ' ' + str(hopred[1, oind]) + ' Naive')
hopreds.append((hopred[0, oind], hopred[1, oind]))
rpool = {}
rpool['Y'] = trainlabels
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hodualrls = GlobalRankRLS(**rpool)
hodualrls.solve(regparam)
hopred_start, hopred_end = hodualrls.leave_pair_out(
np.array([hoindices[0], 1, 1, 2]),
np.array([hoindices[1], 2, 3, 3]))
print(
str(hopred_start[0][1]) + ' ' + str(hopred_end[0][1]) +
' Fast')
hopreds.append((hopred_start[0][1], hopred_end[0][1]))
self.assertAlmostEqual(hopreds[0][0], hopreds[1][0])
self.assertAlmostEqual(hopreds[0][1], hopreds[1][1])
#Test with single output just in case.
rpool = {}
rpool['Y'] = trainlabels[:, 1]
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hodualrls = GlobalRankRLS(**rpool)
hodualrls.solve(regparam)
hopred_start, hopred_end = hodualrls.leave_pair_out(
np.array([hoindices[0], 1, 1, 2]),
np.array([hoindices[1], 2, 3, 3]))
#print(str(hopred_start[0]) + ' ' + str(hopred_end[0]) + ' Fast')
#Test with single output just in case.
rpool = {}
rpool['Y'] = trainlabels[:, 1]
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hodualrls = GlobalRankRLS(**rpool)
hodualrls.solve(regparam)
hopred_start, hopred_end = hodualrls.leave_pair_out(
np.array([hoindices[0]]), np.array([hoindices[1]]))
#print(str(hopred_start) + ' ' + str(hopred_end) + ' Fast')
hopreds = []
rpool = {}
rpool['Y'] = trainlabels
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hoprimalrls = GlobalRankRLS(**rpool)
hoprimalrls.solve(regparam)
hopred = hoprimalrls.holdout(hoindices)
print(str(hopred[0, oind]) + ' ' + str(hopred[1, oind]) + ' HO')
hopreds.append((hopred[0, oind], hopred[1, oind]))
hopred = Xtest * la.inv(Xcv.T * Lcv * Xcv + regparam *
np.mat(np.eye(n))) * Xcv.T * Lcv * Ycv
print(
str(hopred[0, oind]) + ' ' + str(hopred[1, oind]) +
' Dumb (primal)')
hopreds.append((hopred[0, oind], hopred[1, oind]))
hopred0 = hopreds.pop(0)
for hopred in hopreds:
self.assertAlmostEqual(hopred0[0], hopred[0])
self.assertAlmostEqual(hopred0[1], hopred[1])
def testAllPairsRankRLS(self):
print("Testing the cross-validation routines of the GlobalRankRLS module.\n\n")
np.random.seed(100)
floattype = np.float64
m, n, h = 30, 200, 10
Xtrain = np.random.rand(m, n)
trainlabels = np.random.rand(m, h)
trainkm = np.dot(Xtrain, Xtrain.T)
ylen = 1
L = np.mat(m * np.eye(m) - np.ones((m, m), dtype=floattype))
hoindices = [5, 7]
hoindices3 = [5, 7, 9]
hocompl = list(set(range(m)) - set(hoindices))
hocompl3 = list(set(range(m)) - set(hoindices3))
loglambdas = range(-5, 5)
for j in range(0, len(loglambdas)):
regparam = 2. ** loglambdas[j]
print("\nRegparam 2^%1d" % loglambdas[j])
Kcv = trainkm[np.ix_(hocompl, hocompl)]
Ycv = trainlabels[hocompl]
Ktest = trainkm[np.ix_(hocompl, hoindices)]
Xcv = Xtrain[hocompl]
Xtest = Xtrain[hoindices]
Lcv = np.mat((m - 2) * np.eye(m - 2) - np.ones((m - 2, m - 2), dtype=floattype))
oind = 1
rpool = {}
rpool['Y'] = Ycv
rpool['X'] = Xcv
rpool['regparam'] = regparam
naivedualrls = GlobalRankRLS(**rpool)
naivedualrls.solve(regparam)
hopreds = []
hopred = naivedualrls.predictor.predict(Xtest)
print(str(hopred[0, oind]) + ' ' + str(hopred[1, oind]) + ' Naive')
hopreds.append((hopred[0, oind], hopred[1, oind]))
rpool = {}
rpool['Y'] = trainlabels
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hodualrls = GlobalRankRLS(**rpool)
hodualrls.solve(regparam)
hopred_start, hopred_end = hodualrls.leave_pair_out(np.array([hoindices[0], 1, 1, 2]), np.array([hoindices[1], 2, 3, 3]))
print(str(hopred_start[0][1]) + ' ' + str(hopred_end[0][1]) + ' Fast')
hopreds.append((hopred_start[0][1], hopred_end[0][1]))
self.assertAlmostEqual(hopreds[0][0], hopreds[1][0])
self.assertAlmostEqual(hopreds[0][1], hopreds[1][1])
#Test with single output just in case.
rpool = {}
rpool['Y'] = trainlabels[:, 1]
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hodualrls = GlobalRankRLS(**rpool)
hodualrls.solve(regparam)
hopred_start, hopred_end = hodualrls.leave_pair_out(np.array([hoindices[0], 1, 1, 2]), np.array([hoindices[1], 2, 3, 3]))
#print(str(hopred_start[0]) + ' ' + str(hopred_end[0]) + ' Fast')
#Test with single output just in case.
rpool = {}
rpool['Y'] = trainlabels[:, 1]
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hodualrls = GlobalRankRLS(**rpool)
hodualrls.solve(regparam)
hopred_start, hopred_end = hodualrls.leave_pair_out(np.array([hoindices[0]]), np.array([hoindices[1]]))
#print(str(hopred_start) + ' ' + str(hopred_end) + ' Fast')
hopreds = []
rpool = {}
rpool['Y'] = trainlabels
rpool['X'] = Xtrain
rpool['regparam'] = regparam
hoprimalrls = GlobalRankRLS(**rpool)
hoprimalrls.solve(regparam)
hopred = hoprimalrls.holdout(hoindices)
print(str(hopred[0, oind]) + ' ' + str(hopred[1, oind]) + ' HO')
hopreds.append((hopred[0, oind], hopred[1, oind]))
hopred = Xtest * la.inv(Xcv.T * Lcv * Xcv + regparam * np.mat(np.eye(n))) * Xcv.T * Lcv * Ycv
print(str(hopred[0, oind]) + ' ' + str(hopred[1, oind]) + ' Dumb (primal)')
hopreds.append((hopred[0, oind], hopred[1, oind]))
hopred0 = hopreds.pop(0)
for hopred in hopreds:
self.assertAlmostEqual(hopred0[0],hopred[0])
self.assertAlmostEqual(hopred0[1],hopred[1])
