Reformatting with black

Updating scripts and formatting

Author: rosecers

notebooks/1_LinearMethods.ipynb

@@ -43,7 +43,7 @@
     "    table_from_dict,\n",
     "    check_mirrors,\n",
     ")\n",
-    "from skcosmo.preprocessing import KernelFlexibleCenterer as KFC\n",
+    "from skcosmo.preprocessing import KernelNormalizer as KFC\n",
     "\n",
     "cmaps = get_cmaps()\n",
     "plt.style.use(\"../utilities/kernel_pcovr.mplstyle\")\n",
@@ -682,7 +682,7 @@
     "plt.xlabel(\"n\")\n",
     "plt.ylabel(r\"$v_n$\")\n",
     "\n",
-    "np.linalg.norm(v_K[: v_C.shape[0]] - v_C)"
+    "np.linalg.norm(v_K[: min(v_K.shape[0], v_C.shape[0])] - v_C[: min(v_K.shape[0], v_C.shape[0])])"
    ]
   },
   {
@@ -1267,9 +1267,6 @@
    "source": [
     "## Download chemiscope from https://github.com/cosmo-epfl/chemiscope/\n",
     "from chemiscope import write_input\n",
-    "\n",
-    "import ase\n",
-    "from ase import io\n",
     "import json"
    ]
   },
@@ -1279,8 +1276,6 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "frames = ase.io.read(\"../datasets/CSD-1000R.xyz\", \":10\")\n",
-    "\n",
     "# We want to have prediction and projection for all the environments\n",
     "lr = Ridge(alpha=best_regularization)\n",
     "lr.fit(X_train, Y_train)\n",
@@ -1295,21 +1290,21 @@
    ]
   },
   {
-   "cell_type": "code",
-   "execution_count": null,
+   "cell_type": "markdown",
    "metadata": {},
-   "outputs": [],
    "source": [
+    "``` python\n",
+    "\n",
     "data = write_input(\n",
     "    # File to write \n",
     "    path=\"../CSD-1000R.json\",\n",
-    "    \n",
+    "\n",
     "    # Dataset name\n",
     "    meta={\"name\": \"CSD-1000R\"},\n",
-    "    \n",
+    "\n",
     "    # Structures in the dataset\n",
-    "    frames=frames,\n",
-    "    \n",
+    "    frames=# corresponding ASE-type frames,\n",
+    "\n",
     "    # projection of the dataset, and properties predictions\n",
     "    properties={\n",
     "        **{f\"PCA_{i+1}\": {\"target\": \"atom\", \"values\": T[:, i]} for i in range(n_PC)},\n",
@@ -1326,10 +1321,11 @@
     "            \"values\": np.concatenate([frame.numbers for frame in frames])\n",
     "        },\n",
     "    },\n",
-    "    \n",
+    "\n",
     "    # This should be the same cutoff used to generate the features\n",
     "    cutoff=3.5,\n",
-    ")"
+    ")\n",
+    "```"
    ]
   },
   {

notebooks/2_PrincipalCovariatesRegression.ipynb

@@ -39,7 +39,7 @@
     "from sklearn.linear_model import Ridge\n",
     "from sklearn.decomposition import PCA\n",
     "\n",
-    "from skcosmo.pcovr import PCovR\n",
+    "from skcosmo.decomposition import PCovR\n",
     "\n",
     "\n",
     "cmaps = get_cmaps()\n",
@@ -978,7 +978,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from skcosmo.pcovr import PCovR"
+    "from skcosmo.decomposition import PCovR"
    ]
   },
   {

notebooks/3_KernelMethods.ipynb

@@ -45,7 +45,7 @@
     "from sklearn.kernel_ridge import KernelRidge\n",
     "from functools import partial\n",
     "\n",
-    "from skcosmo.pcovr import KPCovR\n",
+    "from skcosmo.decomposition import KPCovR\n",
     "from skcosmo.preprocessing import KernelNormalizer\n",
     "\n",
     "cmaps = get_cmaps()\n",
@@ -1454,7 +1454,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from skcosmo.pcovr import KPCovR"
+    "from skcosmo.decomposition import KPCovR"
    ]
   },
   {

notebooks/4_SparseKernelMethods.ipynb

@@ -35,8 +35,8 @@
     "from sklearn.metrics.pairwise import linear_kernel, rbf_kernel\n",
     "from sklearn.kernel_ridge import KernelRidge\n",
     "from sklearn.decomposition import KernelPCA\n",
-    "from skcosmo.pcovr import KPCovR\n",
-    "from skcosmo.selection import SampleFPS\n",
+    "from skcosmo.decomposition import KPCovR\n",
+    "from skcosmo.sample_selection import FPS\n",
     "from skcosmo.preprocessing import SparseKernelCenterer\n",
     "from functools import partial\n",
     "\n",
@@ -121,8 +121,8 @@
    "source": [
     "n_active = 20\n",
     "\n",
-    "fps_selector = SampleFPS(mixing=1.0, X=X_train)\n",
-    "fps_idxs = fps_selector.select(n_active)\n",
+    "fps_selector = FPS(n_samples_to_select=n_active)\n",
+    "fps_idxs = fps_selector.fit(X_train).selected_idx_\n",
     "\n",
     "Xsparse = X_train[fps_idxs, :]"
    ]

notebooks/5_CUR.ipynb

@@ -45,8 +45,9 @@
     ")\n",
     "from sklearn.decomposition import PCA\n",
     "from sklearn.linear_model import Ridge as LR\n",
+    "from skcosmo.feature_selection import CUR, FPS\n",
     "from skcosmo.selection import FeatureCUR, FeatureFPS\n",
-    "from skcosmo.pcovr import pcovr_covariance\n",
+    "from skcosmo.utils import pcovr_covariance\n",
     "\n",
     "cmaps = get_cmaps()\n",
     "plt.style.use(\"../utilities/kernel_pcovr.mplstyle\")\n",
@@ -458,7 +459,7 @@
    "outputs": [],
    "source": [
     "ns = np.array([int(10 ** x) for x in np.linspace(1, np.log10(nCUR), 20)])\n",
-    "ifps = FeatureFPS(X=X_train).select(n)\n",
+    "ifps = FPS(n_features_to_select=n).fit(X_train).selected_idx_\n",
     "\n",
     "X_CUR = [approx_X(X_train, idxs[:n]) for n in ns]\n",
     "X_FPS = [approx_X(X_train, ifps[:n]) for n in ns]\n",
@@ -1176,8 +1177,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "cur = FeatureCUR(X=X_train, progress_bar=True)\n",
-    "idx_cur = cur.select(max(ns))"
+    "cur = CUR(n_features_to_select=max(ns), progress_bar=True)\n",
+    "idx_cur = cur.fit(X_train).selected_idx_"
    ]
   },
   {

tests/pcovr.py

@@ -4,36 +4,34 @@
 
 from utilities.sklearn_covr.pcovr import PCovR
 
+
 def run_tests():
-    data = np.load('./tests/CSD-test.npz')
+    data = np.load("./tests/CSD-test.npz")
     X = data["X"]
     Y = data["Y"]
 
     # Basic Test of PCovR Errors
     lr_errors = np.nan * np.zeros(21)
     pca_errors = np.nan * np.zeros(21)
-    for i, alpha in enumerate(np.linspace(0,1,21)):
-        pcovr = PCovR(alpha = alpha,
-                      n_components=2,
-                      regularization=1e-6,
-                      tol=1e-12)
+    for i, alpha in enumerate(np.linspace(0, 1, 21)):
+        pcovr = PCovR(alpha=alpha, n_components=2, regularization=1e-6, tol=1e-12)
         pcovr.fit(X, Y)
 
         T = pcovr.transform(X)
         Xr = pcovr.inverse_transform(T)
         Yp = pcovr.predict(X)
-        lr_errors[i] = np.linalg.norm(Y-Yp)**2.0 / np.linalg.norm(Y)**2.0
-        pca_errors[i] = np.linalg.norm(X-Xr)**2.0 / np.linalg.norm(X)**2.0
+        lr_errors[i] = np.linalg.norm(Y - Yp) ** 2.0 / np.linalg.norm(Y) ** 2.0
+        pca_errors[i] = np.linalg.norm(X - Xr) ** 2.0 / np.linalg.norm(X) ** 2.0
 
         assert not np.isnan(lr_errors[i]) and not np.isnan(pca_errors[i])
 
-    assert all(lr_errors[i] <= lr_errors[i+1] and pca_errors[i] >= pca_errors[i+1]for i in range(len(lr_errors)-1))
+    assert all(
+        lr_errors[i] <= lr_errors[i + 1] and pca_errors[i] >= pca_errors[i + 1]
+        for i in range(len(lr_errors) - 1)
+    )
 
     # Test of PCovR Fitting
-    pcovr = PCovR(alpha = 0.5,
-                  n_components=2,
-                  regularization=1e-6,
-                  tol=1e-12)
+    pcovr = PCovR(alpha=0.5, n_components=2, regularization=1e-6, tol=1e-12)
 
     try:
         T = pcovr.transform(X)
@@ -47,7 +45,6 @@ def run_tests():
     assert check_X_y(X, T, multi_output=True)
 
 
-
 if __name__ == "__main__":
     run_tests()
     print("Everything passed")