Fixed kernel test

Author: charnley

Makefile

@@ -22,10 +22,29 @@ format:
 
 test:
 	${python} -m pytest -rs \
-	./tests/test_kernels.py \
-	./tests/test_solvers.py \
-	./tests/test_distance.py \
-	./tests/test_slatm.py
+	tests/test_acsf_linear_angles.py \
+	tests/test_acsf.py \
+	tests/test_arad.py \
+	tests/test_armp.py \
+	tests/test_compound.py \
+	tests/test_distance.py \
+	tests/test_energy_krr_atomic_cmat.py \
+	tests/test_energy_krr_bob.py \
+	tests/test_energy_krr_cmat.py \
+	tests/test_fchl_acsf_energy.py \
+	tests/test_fchl_acsf_forces.py \
+	tests/test_fchl_acsf.py \
+	tests/test_fchl_electric_field.py \
+	tests/test_fchl_force.py \
+	tests/test_fchl_scalar.py \
+	tests/test_kernel_derivatives.py \
+	tests/test_kernels.py \
+	tests/test_mrmp.py \
+	tests/test_neural_network.py \
+	tests/test_representations.py \
+	tests/test_slatm.py \
+	tests/test_solvers.py \
+	tests/test_symm_funct.py
 
 types:
 	${python} -m monkeytype run $(which pytest) ./tests/

src/qmllib/representations/slatm.py

@@ -1,9 +1,7 @@
-from __future__ import print_function
-
 import itertools as itl
 
 import numpy as np
-import scipy.spatial.distance as ssd
+import scipy.spatial.distance as spatial_distance
 
 from .fslatm import fget_sbop, fget_sbop_local, fget_sbot, fget_sbot_local
 
@@ -16,7 +14,7 @@ def update_m(obj, ia, rcut=9.0, pbc=None):
 
     zs, coords, c = obj
     v1, v2, v3 = c
-    vs = ssd.norm(c, axis=0)
+    vs = spatial_distance.norm(c, axis=0)
 
     nns = []
     for i, vi in enumerate(vs):
@@ -68,7 +66,7 @@ def update_m(obj, ia, rcut=9.0, pbc=None):
     if na == 1:
         ds = np.array([[0.0]])
     else:
-        ds = ssd.squareform(ssd.pdist(coords))
+        ds = spatial_distance.squareform(spatial_distance.pdist(coords))
 
     zs_u = []
     coords_u = []
@@ -86,7 +84,7 @@ def update_m(obj, ia, rcut=9.0, pbc=None):
                 ts[iau] = np.dot(n123s[iau], c)
 
             coords_iu = coords[i] + ts  # np.dot(n123s, c)
-            dsi = ssd.norm(coords_iu - cia, axis=1)
+            dsi = spatial_distance.norm(coords_iu - cia, axis=1)
             filt = np.logical_and(dsi > 0, dsi <= rcut)
             nx = filt.sum()
             zs_u += [

tests/test_kernels.py

@@ -1,11 +1,10 @@
-import os
+from pathlib import Path
 
 import numpy as np
-import pytest
+from conftest import ASSETS
 from scipy.stats import wasserstein_distance
 from sklearn.decomposition import KernelPCA
 
-import qmllib
 from qmllib.kernels import (
     gaussian_kernel,
     gaussian_kernel_symmetric,
@@ -17,14 +16,15 @@
     sargan_kernel,
     wasserstein_kernel,
 )
+from qmllib.representations import generate_bob
+from qmllib.utils.xyz_format import read_xyz
 
 
-def get_energies(filename):
+def get_energies(filename: Path):
     """Returns a dictionary with heats of formation for each xyz-file."""
 
-    f = open(filename, "r")
-    lines = f.readlines()
-    f.close()
+    with open(filename, "r") as f:
+        lines = f.readlines()
 
     energies = dict()
 
@@ -239,16 +239,10 @@ def array_nan_close(a, b):
     return np.allclose(a[m], b[m], atol=1e-8, rtol=0.0)
 
 
-@pytest.mark.skip(reason="Removing all Compound classes")
 def test_kpca():
 
-    test_dir = os.path.dirname(os.path.realpath(__file__))
-
     # Parse file containing PBE0/def2-TZVP heats of formation and xyz filenam
-    data = get_energies(test_dir + "/data/hof_qm7.txt")
-
-    # Generate a list of qmllib.Compound() objects
-    mols = []
+    data = get_energies(ASSETS / "hof_qm7.txt")
 
     keys = sorted(data.keys())
 
@@ -257,17 +251,24 @@ def test_kpca():
 
     n_mols = 100
 
+    representations = []
+
     for xyz_file in keys[:n_mols]:
 
-        mol = qmllib.Compound(xyz=test_dir + "/qm7/" + xyz_file)
-        mol.properties = data[xyz_file]
-        mol.generate_bob()
-        mols.append(mol)
+        filename = ASSETS / "qm7" / xyz_file
+        coordinates, atoms = read_xyz(filename)
 
-    X = np.array([mol.representation for mol in mols])
+        atomtypes = np.unique(atoms)
+        representation = generate_bob(atoms, coordinates, atomtypes)
+        representations.append(representation)
+
+    X = np.array([representation for representation in representations])
     K = laplacian_kernel(X, X, 2e5)
 
+    # calculate pca
     pcas_qml = kpca(K, n=10)
+
+    # Calculate with sklearn
     pcas_sklearn = KernelPCA(10, eigen_solver="dense", kernel="precomputed").fit_transform(K)
 
     assert array_nan_close(