def cv_count_test():
"cv index iterator test for proper count and uniqueness"
msg1 = "Each element was not exactly once in test when n=%d and kfold=%d"
msg2 = "Test elements are not unique when n=%d and kfold=%d"
msg3 = "Train elements are not unique when n=%d and kfold=%d"
msg4 = "Each element was not in train the expected number of times "
msg4 += "when n=%d and kfold=%d"
msg5 = "Same element in train and test when n=%d and kfold=%d"
msg6 = "Min index value is not 0 when n=%d and kfold=%d"
msg7 = "Max index value is not n-1 when n=%d and kfold=%d"
msg8 = "test index is empty when n=%d and kfold=%d"
for n in range(2, 8):
for k in range(2, n + 1):
trains = []
tests = []
for train, test in cross_validation(n, k):
trains += train
tests += test
yield assert_, len(test) > 0, msg8 % (n, k)
ncommon = len(set(train) & set(test))
yield assert_equal, ncommon, 0, msg5 % (n, k)
unique = set(train) | set(test)
yield assert_equal, min(unique), 0, msg6 % (n, k)
yield assert_equal, max(unique), n - 1, msg7 % (n, k)
yield assert_equal, len(tests), n, msg1 % (n, k)
yield assert_equal, len(set(tests)), n, msg2 % (n, k)
yield assert_equal, len(set(trains)), n, msg3 % (n, k)
yield assert_equal, len(trains), n * k - n, msg4 % (n, k)
def cv_count_test():
"cv index iterator test for proper count and uniqueness"
msg1 = "Each element was not exactly once in test when n=%d and kfold=%d"
msg2 = "Test elements are not unique when n=%d and kfold=%d"
msg3 = "Train elements are not unique when n=%d and kfold=%d"
msg4 = "Each element was not in train the expected number of times "
msg4 += "when n=%d and kfold=%d"
msg5 = "Same element in train and test when n=%d and kfold=%d"
msg6 = "Min index value is not 0 when n=%d and kfold=%d"
msg7 = "Max index value is not n-1 when n=%d and kfold=%d"
msg8 = "test index is empty when n=%d and kfold=%d"
for n in range(2, 8):
for k in range(2, n+1):
trains = []
tests = []
for train, test in cross_validation(n, k):
trains += train
tests += test
yield assert_, len(test) > 0, msg8 % (n, k)
ncommon = len(set(train) & set(test))
yield assert_equal, ncommon, 0, msg5 % (n, k)
unique = set(train) | set(test)
yield assert_equal, min(unique), 0, msg6 % (n, k)
yield assert_equal, max(unique), n-1, msg7 % (n, k)
yield assert_equal, len(tests), n, msg1 % (n, k)
yield assert_equal, len(set(tests)), n, msg2 % (n, k)
yield assert_equal, len(set(trains)), n, msg3 % (n, k)
yield assert_equal, len(trains), n*k - n, msg4 % (n, k)
def cv_repeatability_test():
"cv index iterator test for repeatability"
msg = "%s indices were not repeatable when n=%d and kfold=%d"
n = 5
k = 3
train1 = []
test1 = []
shuffle = np.random.RandomState([1, 2, 3]).shuffle
for train, test in cross_validation(n, k, shuffle):
train1 += train
test1 += test
train2 = []
test2 = []
shuffle = np.random.RandomState([1, 2, 3]).shuffle
for train, test in cross_validation(n, k, shuffle):
train2 += train
test2 += test
yield assert_equal, train1, train2, msg % ('train', n, k)
def cv_repeatability_test():
"cv index iterator test for repeatability"
msg = "%s indices were not repeatable when n=%d and kfold=%d"
n = 5
k = 3
train1 = []
test1 = []
shuffle = np.random.RandomState([1, 2, 3]).shuffle
for train, test in cross_validation(n, k, shuffle):
train1 += train
test1 += test
train2 = []
test2 = []
shuffle = np.random.RandomState([1, 2, 3]).shuffle
for train, test in cross_validation(n, k, shuffle):
train2 += train
test2 += test
yield assert_equal, train1, train2, msg % ('train', n, k)
def split_test():
"split test"
ys = [la.rand(3), la.rand(3, 4), la.rand(2, 3, 4)]
msg = 'Failed on shape %s, axis %s'
for y in ys:
for axis in range(y.ndim):
cv = cross_validation(y.shape[axis], y.shape[axis])
for idx_train, idx_test in cv:
ytrain, ytest = split(y, idx_train, idx_test, axis)
y2 = ytrain.merge(ytest)
errmsg = msg % (str(y.shape), str(axis))
yield assert_larry_equal, y, y2, errmsg, y
def split_test():
"split test"
ys = [la.rand(3), la.rand(3,4), la.rand(2,3,4)]
msg = 'Failed on shape %s, axis %s'
for y in ys:
for axis in range(y.ndim):
cv = cross_validation(y.shape[axis], y.shape[axis])
for idx_train, idx_test in cv:
ytrain, ytest = split(y, idx_train, idx_test, axis)
y2 = ytrain.merge(ytest)
errmsg = msg % (str(y.shape), str(axis))
yield assert_larry_equal, y, y2, errmsg, y
