| 1566 | ) |
| 1567 | |
| 1568 | def test_on_transformed(self): |
| 1569 | n_samples = 10 |
| 1570 | n_features = 5 |
| 1571 | query_strategies = [ |
| 1572 | # TODO remove, added just to make sure on_transformed doesn't break anything |
| 1573 | # but it has no influence on this strategy, nothing special tested here |
| 1574 | mock.MockFunction(return_val=[np.random.randint(0, n_samples)]) |
| 1575 | |
| 1576 | # add further strategies which work with instance representations |
| 1577 | # no further ones as of 25.09.2020 |
| 1578 | ] |
| 1579 | X_pool = np.random.rand(n_samples, n_features) |
| 1580 | |
| 1581 | # use pandas data frame as X_pool, which will be transformed back to numpy with sklearn pipeline |
| 1582 | X_pool = pd.DataFrame(X_pool) |
| 1583 | |
| 1584 | y_pool = np.random.rand(n_samples) |
| 1585 | train_idx = np.random.choice(range(n_samples), size=2, replace=False) |
| 1586 | |
| 1587 | learner_list = [modAL.models.learners.ActiveLearner( |
| 1588 | estimator=make_pipeline( |
| 1589 | FunctionTransformer(func=pd.DataFrame.to_numpy), |
| 1590 | GaussianProcessRegressor() |
| 1591 | ), |
| 1592 | # committee learners can contain different amounts of |
| 1593 | # different instances |
| 1594 | X_training=X_pool.iloc[train_idx[( |
| 1595 | np.arange(i + 1) + i) % len(train_idx)]], |
| 1596 | y_training=y_pool[train_idx[( |
| 1597 | np.arange(i + 1) + i) % len(train_idx)]], |
| 1598 | ) for i in range(3)] |
| 1599 | |
| 1600 | for query_strategy in query_strategies: |
| 1601 | committee = modAL.models.learners.CommitteeRegressor( |
| 1602 | learner_list=learner_list, |
| 1603 | query_strategy=query_strategy, |
| 1604 | on_transformed=True |
| 1605 | ) |
| 1606 | query_idx, query_inst = committee.query(X_pool) |
| 1607 | committee.teach(X_pool.iloc[query_idx], y_pool[query_idx]) |
| 1608 | |
| 1609 | |
| 1610 | class TestMultilabel(unittest.TestCase): |