def test_double_fault():
labels = np.array([0, 0, 0, 0, 1, 1, 1])
pred1 = np.array([1, 0, 1, 0, 0, 0, 0])
pred2 = np.array([1, 0, 0, 0, 1, 0, 0])
actual = double_fault(labels, pred1, pred2)
assert actual == 3. / 7 # three common errors out of 7 predictions
def diversity_matrix(pool_classifiers, X, y, pruned_indices):
if pruned_indices is not None:
preds = np.array([
pool_classifiers.estimators_[id].predict(X)
for id in pruned_indices
])
else:
preds = np.array([
estimator.predict(X) for estimator in pool_classifiers.estimators_
])
n = len(preds)
div = np.zeros((n, n))
for i in range(n):
div[i][i] = double_fault(y, preds[i], preds[i])
for i in range(n):
for j in range(i + 1, n):
div[i][j] = div[j][i] = double_fault(y, preds[i], preds[j])
return div
def test_double_fault_ones():
df = double_fault(y_dsel_ex1, y_pred_ones, y_pred_ones)
assert df == 9.0 / 15.0
def test_double_fault_zeros():
df = double_fault(y_dsel_ex1, y_pred_zeros, y_pred_zeros)
assert df == 6.0 / 15.0
def test_double_fault_order():
df1 = double_fault(y_dsel_ex1, y_pred_ones, y_pred_zeros)
df2 = double_fault(y_dsel_ex1, y_pred_zeros, y_pred_ones)
assert df1 == df2
def test_double_fault_ones_zeros():
df = double_fault(y_dsel_ex1, y_pred_ones, y_pred_zeros)
assert df == 0.0
def test_double_fault_ones(example_diversity_ones_zeros):
y, y_pred_ones, y_pred_zeros = example_diversity_ones_zeros
df = double_fault(y, y_pred_ones, y_pred_ones)
assert df == 9.0 / 15.0
def test_double_fault_order(example_diversity_ones_zeros):
y, y_pred_ones, y_pred_zeros = example_diversity_ones_zeros
df1 = double_fault(y, y_pred_ones, y_pred_zeros)
df2 = double_fault(y, y_pred_zeros, y_pred_ones)
assert df1 == df2
