test: add functional tests for synthetic experiments

experiments/synthetic/clustering.py

@@ -31,17 +31,15 @@
 warnings.filterwarnings("ignore")
 
 
-if __name__ == "__main__":
-    arguments = docopt(__doc__)
-
-    SEED = int(arguments['--seed'])
+def run(args: dict) -> None:
+    """Main runner."""
+    SEED = int(args['--seed'])
     set_seed(SEED)
 
-    x, y = load(str(arguments['<d>']))
-
+    x, y = load(str(args['<d>']))
 
     dm = get_dip_means(
-        arguments=arguments,
+        arguments=args,
         seed=SEED
     )
 
@@ -50,3 +48,9 @@
     logger.info(f'The NMI score is {normalized_mutual_info_score(y, clusters)}')
 
     plot_clustered_data(x, clusters)
+
+
+if __name__ == "__main__":
+    arguments = docopt(__doc__)
+
+    run(args=arguments)

experiments/synthetic/two_gaussians_mix.py

@@ -31,24 +31,24 @@
 warnings.filterwarnings("ignore")
 
 
-if __name__ == "__main__":
-    arguments = docopt(__doc__)
+def run(args: dict) -> None:
+    """Main runner."""
 
-    SEED = int(arguments['--seed'])
+    SEED = int(args['--seed'])
     set_seed(SEED)
 
-    n_samples = int(arguments['--samples'])
-    std = float(arguments['--noise'])
+    n_samples = int(args['--samples'])
+    std = float(args['--noise'])
     g = TwoDimGaussianSumGenerator(
       n=n_samples,
       cluster_std=std, 
       random_state=SEED
     )
 
-    mct = get_monte_carlo_test(arguments=arguments, workers_num=1)
+    mct = get_monte_carlo_test(arguments=args, workers_num=1)
 
     tr = 'unimodal' if mct.test(g.x) else 'bimodal'
-    msg = dict(arguments)
+    msg = dict(args)
     msg['groundtruth'] = g.t
     msg['result'] = tr
     msg.pop('--help')
@@ -63,3 +63,9 @@
         'The inputs and the output of the experiments is: '
         f'{msg}'
     )
+
+
+if __name__ == "__main__":
+    arguments = docopt(__doc__)
+
+    run(args=arguments)

experiments/synthetic/unimodality.py

@@ -51,20 +51,20 @@ def get_dataset(name: str) -> Callable:
         raise ValueError(msg)
 
 
-if __name__ == "__main__":
-    arguments = docopt(__doc__)
+def run(args: dict) -> None:
+    """Main runner."""
 
-    SEED = int(arguments['--seed'])
+    SEED = int(args['--seed'])
     set_seed(SEED)
 
-    data_func = get_dataset(str(arguments['<d>']))
-    n_samples = int(arguments['--samples'])
-    noise = float(arguments['--noise'])
+    data_func = get_dataset(str(args['<d>']))
+    n_samples = int(args['--samples'])
+    noise = float(args['--noise'])
     x, y = data_func(n_samples=n_samples, noise=noise, random_state=SEED)
 
-    mct = get_monte_carlo_test(arguments=arguments, workers_num=1)
+    mct = get_monte_carlo_test(arguments=args, workers_num=1)
 
-    msg = dict(arguments)
+    msg = dict(args)
     msg['result'] = 'unimodal' if mct.test(x) else 'multimodal'
     msg.pop('--help')
 
@@ -78,3 +78,8 @@ def get_dataset(name: str) -> Callable:
         f'{msg}'
     )
 
+
+if __name__ == "__main__":
+    arguments = docopt(__doc__)
+
+    run(args=arguments)

tests/experiments/test_synthetic.py

@@ -0,0 +1,68 @@
+"""Tests written for the synthetic experiments module."""
+import pytest
+
+from experiments.common import get_dip_means
+from experiments.synthetic.two_gaussians_mix import run as run_gaussians
+from experiments.synthetic.unimodality import run as run_unimodality
+
+
+@pytest.mark.parametrize("projector", ['jl', 'i'])
+@pytest.mark.parametrize("distance", ['euclidean', 'mahalanobis'])
+@pytest.mark.parametrize("observer", ['percentile', 'random'])
+def test_get_dip_means(projector: str, distance: str, observer: str) -> None:
+    """Test that the factory for DipMeans works properly."""
+
+    args = {
+        '<pj>': projector,
+        '--dist': distance,
+        '--obs': observer,
+        '<pv>': 0.05,
+        '<sims>': 5,
+    }
+
+    get_dip_means(args, seed=42)
+
+
+@pytest.mark.parametrize("projector", ['jl', 'i'])
+@pytest.mark.parametrize("distance", ['euclidean', 'mahalanobis'])
+@pytest.mark.parametrize("observer", ['percentile', 'random'])
+def test_run_gaussians(projector: str, distance: str, observer: str) -> None:
+    """Test that the unimodality experiments on 2D gaussians mixture work properly."""
+
+    args = {
+        '--seed': 42,
+        '--samples': 200,
+        '--noise': 0.5,
+        '<pj>': projector,
+        '--dist': distance,
+        '--obs': observer,
+        '<pv>': 0.05,
+        '<sims>': 5,
+        '--help': False,
+        '--plot': 'False',
+    }
+
+    run_gaussians(args)
+
+
+@pytest.mark.parametrize("projector", ['jl', 'i'])
+@pytest.mark.parametrize("distance", ['euclidean', 'mahalanobis'])
+@pytest.mark.parametrize("observer", ['percentile', 'random'])
+def test_run_unimodality(projector: str, distance: str, observer: str) -> None:
+    """Test that the unimodality experiments work properly."""
+
+    args = {
+        '--seed': 42,
+        '<d>': 'circles',
+        '--samples': 200,
+        '--noise': 0.5,
+        '<pj>': projector,
+        '--dist': distance,
+        '--obs': observer,
+        '<pv>': 0.05,
+        '<sims>': 5,
+        '--help': False,
+        '--plot': 'False',
+    }
+
+    run_unimodality(args)