Add Ridge and StandardScaler

.gitignore

@@ -1,57 +1,16 @@
-*.so
-/*.egg-info
-*.pyc
-__pycache__
-*.DS_store
-*.npz
-*.dat
-*.npy
-*.xvg
-*.egg-info
-.duecredit.p
 .tox/
-.swp
-*.npz
-*.dat
-*.npy
-*.xvg
-*.lock
-
-# Cython generated files
-_cutil.c
+build/
+dist/
+*.egg-info
+__pycache__/
 
 # Unit test / coverage reports
-htmlcov/
-.tox/
 .coverage
 .coverage.*
-.cache
 nosetests.xml
 coverage.xml
 *.cover
 .hypothesis/
-.pytest_cache/
-
-# Sphinx documentation
-docs/_build/
-docs/html
-docs/latex
-
-# editors and IDEs
-*.sw[a-z]
-*~
-.idea
-.vscode
-
-# Ignore build dir
-/build
-/dist
-
-# Ignore the .DS_Store file in the osx file system
-*.DS_store
-
-# duecredit
-.duecredit.p
 
 # editors and IDEs
 *.sw[a-z]

examples/README.rst

@@ -1,4 +1,4 @@
 Equisolve Examples
 ==================
 
-This folder consists of introductory examples and examples.
+This folder consists of introductory examples.

examples/linear-model.py

@@ -17,12 +17,10 @@
 import ase.io
 import numpy as np
 from equistore import Labels
-from equistore.operations import slice, sum_over_samples
-from equistore.status import EquistoreError
+from equistore.operations import ones_like, slice, sum_over_samples
 from rascaline import SoapPowerSpectrum
 
 from equisolve.numpy.models.linear_model import Ridge
-from equisolve.utils import dictionary_to_tensormap
 from equisolve.utils.convert import ase_to_tensormap
 
 
@@ -81,23 +79,15 @@
 # rascaline.
 #
 # We now move all keys into properties to access them for our model.
-try:
-    descriptor = descriptor.keys_to_samples("species_center")
-except EquistoreError:
-    pass
-
-try:
-    descriptor = descriptor.keys_to_properties(
-        ["species_neighbor_1", "species_neighbor_2"]
-    )
-except EquistoreError:
-    pass
+
+descriptor = descriptor.keys_to_samples("species_center")
+descriptor = descriptor.keys_to_properties(["species_neighbor_1", "species_neighbor_2"])
 
 # %%
 #
 # The descriptor contains a represenantion with respect to each central atoms per
 # structure. However, our energies as target data is per structure only.
-# Therefore, we sum the features of each center atom per structure.
+# Therefore, we sum the properties of each center atom per structure.
 
 X = sum_over_samples(descriptor, ["center", "species_center"])
 
@@ -109,12 +99,12 @@
 
 # %%
 #
-# As well as 1800 features and 20 sample.
+# As well as 1800 properties and 20 sample.
 #
 # We acces the data using the :meth:``equistore.TensorMap.block`` method
 
 
-print(f"X contains {len(X.block().properties)} features.")
+print(f"X contains {len(X.block().properties)} properties.")
 print(f"X contains {len(X.block().samples)} samples.")
 
 # Target data
@@ -141,9 +131,8 @@
 #
 # For this we create a TensorMap containing with the desired regulerizer
 
-
-alpha_dict = {"values": 1e-5}
-alpha = dictionary_to_tensormap(alpha_dict, X)
+alpha = ones_like(X)
+alpha.block().values[:] *= 1e-5
 
 # %%
 #
@@ -158,15 +147,16 @@
     names=["structure"],
     values=np.array([(0,)]),
 )
+
 alpha = slice(alpha, samples=samples)
 
 # %%
 #
-# In our regulerizer we use the same values for all features. However,
+# In our regulerizer we use the same values for all properties. However,
 # :class:`equisolve.numpy.models.linear_model.Ridge` can also handle different
-# regularization for each feature. You can apply a feature wise regularization by
+# regularization for each property. You can apply a property wise regularization by
 # setting ``"values"`` of ``alpha_dict`` with an 1d array of the same length as the
-# number of features in the training data X (here 7200)
+# number of properties in the training data X (here 7200)
 #
 # With a valid regulerizer object we now initilize the Ridge object.
 # ``parameter_keys`` determines with respect to which parameters the regression is
@@ -181,8 +171,8 @@
 # Next we create a sample weighting :class:`equistiore.TensorMap` that weights energies
 # five times more then the forces.
 
-sw_dict = {"values": 5, "positions": 1}
-sw = dictionary_to_tensormap(sw_dict, y)
+sw = ones_like(y)
+sw.block().values[:] *= 5
 
 # %%
 #
@@ -194,11 +184,12 @@
 
 # Finally we can fit the model using the sample weights defined above.
 
-clf.fit(X, y, sample_weights=sw)
+clf.fit(X, y, sample_weight=sw)
 
 
 # Finally we can predict values and calculate the root mean squre error
 # of our model.
 
 clf.predict(X)
-print(f"RMSE = {clf.score(X, y)}")
+print(f"RMSE energies = {clf.score(X, y, parameter_key='values')[0]:.3f} eV")
+print(f"RMSE forces = {clf.score(X, y, parameter_key='positions')[0]:.3f} eV/Å")

setup.cfg

@@ -46,7 +46,7 @@ packages = find:
 include_package_data = True
 
 install_requires =
-    #equistore @ https://github.com/lab-cosmo/equistore/archive/6ca7fa3.zip
+    equistore @ https://github.com/lab-cosmo/equistore/archive/467baf0.zip
     numpy>=1.19.0
     scipy>=1.0.0