Merge pull request #1508 from helmholtz-analytics/features/1448-rando…

…m_arrays_of_arbitrary_size

Features/1448 Refactor random module

heat/cluster/_kcluster.py

@@ -109,34 +109,9 @@ def _initialize_cluster_centers(self, x: DNDarray):
 
         # initialize the centroids by randomly picking some of the points
         if self.init == "random":
-            # Samples will be equally distributed drawn from all involved processes
-            _, displ, _ = x.comm.counts_displs_shape(shape=x.shape, axis=0)
-            centroids = ht.empty(
-                (self.n_clusters, x.shape[1]), split=None, device=x.device, comm=x.comm
-            )
-            if x.split is None or x.split == 0:
-                for i in range(self.n_clusters):
-                    samplerange = (
-                        x.gshape[0] // self.n_clusters * i,
-                        x.gshape[0] // self.n_clusters * (i + 1),
-                    )
-                    sample = ht.random.randint(samplerange[0], samplerange[1]).item()
-                    proc = 0
-                    for p in range(x.comm.size):
-                        if displ[p] > sample:
-                            break
-                        proc = p
-                    xi = ht.zeros(x.shape[1], dtype=x.dtype)
-                    if x.comm.rank == proc:
-                        idx = sample - displ[proc]
-                        xi = ht.array(x.lloc[idx, :], device=x.device, comm=x.comm)
-                    xi.comm.Bcast(xi, root=proc)
-                    centroids[i, :] = xi
-
-            else:
-                raise NotImplementedError("Not implemented for other splitting-axes")
-
-            self._cluster_centers = centroids
+            idx = ht.random.randint(0, x.shape[0] - 1, size=(self.n_clusters,), split=None)
+            centroids = x[idx, :]
+            self._cluster_centers = centroids if x.split == 1 else centroids.resplit_(None)
 
         # directly passed centroids
         elif isinstance(self.init, DNDarray):
@@ -172,7 +147,7 @@ def _initialize_cluster_centers(self, x: DNDarray):
                     D2 = distances.min(axis=1)
                     D2.resplit_(axis=None)
                     prob = D2 / D2.sum()
-                    random_position = ht.random.rand().item()
+                    random_position = ht.random.rand()
                     sample = 0
                     sum = 0
                     for j in range(len(prob)):

heat/cluster/kmeans.py

@@ -119,7 +119,6 @@ def fit(self, x: DNDarray) -> self:
         # initialize the clustering
         self._initialize_cluster_centers(x)
         self._n_iter = 0
-        matching_centroids = ht.zeros((x.shape[0]), split=x.split, device=x.device, comm=x.comm)
 
         # iteratively fit the points to the centroids
         for epoch in range(self.max_iter):

heat/cluster/kmedians.py

@@ -119,7 +119,7 @@ def fit(self, x: DNDarray):
         # initialize the clustering
         self._initialize_cluster_centers(x)
         self._n_iter = 0
-        matching_centroids = ht.zeros((x.shape[0]), split=x.split, device=x.device, comm=x.comm)
+
         # iteratively fit the points to the centroids
         for epoch in range(self.max_iter):
             # increment the iteration count

heat/cluster/kmedoids.py

@@ -130,7 +130,7 @@ def fit(self, x: DNDarray):
         # initialize the clustering
         self._initialize_cluster_centers(x)
         self._n_iter = 0
-        matching_centroids = ht.zeros((x.shape[0]), split=x.split, device=x.device, comm=x.comm)
+
         # iteratively fit the points to the centroids
         for epoch in range(self.max_iter):
             # increment the iteration count

heat/core/linalg/tests/test_basics.py

@@ -302,13 +302,13 @@ def test_inv(self):
         ainv = ht.linalg.inv(a)
         i = ht.eye(a.shape, split=1, dtype=a.dtype)
         # loss of precision in distributed floating-point ops
-        self.assertTrue(ht.allclose(a @ ainv, i, atol=1e-12))
+        self.assertTrue(ht.allclose(a @ ainv, i, atol=1e-10))
 
         ht.random.seed(42)
         a = ht.random.random((20, 20), dtype=ht.float64, split=0)
         ainv = ht.linalg.inv(a)
         i = ht.eye(a.shape, split=0, dtype=a.dtype)
-        self.assertTrue(ht.allclose(a @ ainv, i, atol=1e-12))
+        self.assertTrue(ht.allclose(a @ ainv, i, atol=1e-10))
 
         with self.assertRaises(RuntimeError):
             ht.linalg.inv(ht.array([1, 2, 3], split=0))

heat/core/linalg/tests/test_svdtools.py

@@ -60,7 +60,7 @@ def test_hsvd_rank_part1(self):
                         )
                         self.assertTrue(V_orth_err <= dtype_tol)
                     true_rel_err = ht.norm(U @ ht.diag(sigma) @ V.T - A) / ht.norm(A)
-                    self.assertTrue(true_rel_err <= err_est)
+                    self.assertTrue(true_rel_err <= err_est or true_rel_err < dtype_tol)
                 else:
                     self.assertEqual(hsvd_rk, 1)
                     self.assertEqual(ht.norm(U), 0)
@@ -96,7 +96,7 @@ def test_hsvd_rank_part1(self):
                         )
                         self.assertTrue(V_orth_err <= dtype_tol)
                     true_rel_err = ht.norm(U @ ht.diag(sigma) @ V.T - A) / ht.norm(A)
-                    self.assertTrue(true_rel_err <= err_est)
+                    self.assertTrue(true_rel_err <= err_est or true_rel_err < dtype_tol)
                     self.assertTrue(true_rel_err <= tol)
                 else:
                     self.assertEqual(hsvd_rk, 1)