Add option for numba acceleration. Improve tests

[MattScicluna]

No description provided.

[bjoaofelipe]

Ran into two errors:

======================================================================
ERROR: test_random_landmarking.test_random_landmarking_with_non_euclidean_distances
Test random landmarking with non-euclidean distance metrics

Traceback (most recent call last):
File "/home/runner/work/graphtools/graphtools/test/test_random_landmarking.py", line 383, in test_random_landmarking_with_non_euclidean_distances
warnings.warn(
UserWarning: Random landmarking may not be respecting the distance parameter. All distance metrics produced identical cluster assignments.

======================================================================
FAIL: test_landmark.test_set_params

Traceback (most recent call last):
File "/home/runner/work/graphtools/graphtools/test/test_landmark.py", line 205, in test_set_params
assert G.get_params() == {
^^^^^^^^^^^^^^^^^^^
AssertionError

---

[bjoaofelipe]

The distance parameter is not being used. Need to start passing it downstream. Every algorithm uses the default euclidean distance

[bjoaofelipe]

Found the error on:

    def build_landmark_op(self):
        """Build the landmark operator

            Calculates spectral clusters on the kernel, and calculates transition
            probabilities between cluster centers by using transition probabilities
            between samples assigned to each cluster.
            
            random_landmarking:
            This method randomly selects n_landmark points and assigns each sample to its nearest landmark
            using Euclidean distance .

        """
        if self.random_landmarking :
            with _logger.log_task("landmark operator"):
                is_sparse = sparse.issparse(self.kernel)
                n_samples = self.data.shape[0]
                rng = np.random.default_rng(self.random_state)
                landmark_indices = rng.choice(n_samples, self.n_landmark, replace=False)
                data = self.data if not hasattr(self, 'data_nu') else self.data_nu # because of the scaling to review
                if n_samples > 5000:   # sklearn.euclidean_distances is faster than cdist for big dataset 
                    distances = euclidean_distances(data, data[landmark_indices])
                else:
                    distances = cdist(data, data[landmark_indices], metric="euclidean")
                self._clusters = np.argmin(distances, axis=1)

We are forcing euclidean distances here, which is not the expected behavior

[MattScicluna]

Found the error on:

    def build_landmark_op(self):
        """Build the landmark operator

            Calculates spectral clusters on the kernel, and calculates transition
            probabilities between cluster centers by using transition probabilities
            between samples assigned to each cluster.
            
            random_landmarking:
            This method randomly selects n_landmark points and assigns each sample to its nearest landmark
            using Euclidean distance .

        """
        if self.random_landmarking :
            with _logger.log_task("landmark operator"):
                is_sparse = sparse.issparse(self.kernel)
                n_samples = self.data.shape[0]
                rng = np.random.default_rng(self.random_state)
                landmark_indices = rng.choice(n_samples, self.n_landmark, replace=False)
                data = self.data if not hasattr(self, 'data_nu') else self.data_nu # because of the scaling to review
                if n_samples > 5000:   # sklearn.euclidean_distances is faster than cdist for big dataset 
                    distances = euclidean_distances(data, data[landmark_indices])
                else:
                    distances = cdist(data, data[landmark_indices], metric="euclidean")
                self._clusters = np.argmin(distances, axis=1)

We are forcing euclidean distances here, which is not the expected behavior

I added a test to test_landmark 04256c3 to confirm that this unexpected behaviour happens for both random landmarking and the default.

[bjoaofelipe]

I will create a new isue mentioning that distances are being ignored by the build_graph class.