Improvements on kernel method and general documentation

- models/kernelridge.py: avoid computing redundant elements when building the covariance matrix. - General improvements to the documentation of the package.
muammar · Aug 13, 2019 · bf88896 · bf88896
1 parent dff38ab
commit bf88896
Show file tree

Hide file tree

Showing 9 changed files with 196 additions and 104 deletions.
diff --git a/docs/environment.yml b/docs/environment.yml
@@ -5,7 +5,7 @@ channels:
 
 dependencies:
     - python==3.7
-    - torch
+    - pytorch
     - pip:
         - matplotlib
         - pyyaml

diff --git a/docs/source/ml4chem.backends.rst b/docs/source/ml4chem.backends.rst
@@ -8,23 +8,23 @@ ml4chem.backends.available module
 ---------------------------------
 
 .. automodule:: ml4chem.backends.available
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.backends.operations module
 ----------------------------------
 
 .. automodule:: ml4chem.backends.operations
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 
 Module contents
 ---------------
 
 .. automodule:: ml4chem.backends
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/source/ml4chem.data.rst b/docs/source/ml4chem.data.rst
@@ -8,39 +8,55 @@ ml4chem.data.handler module
 ---------------------------
 
 .. automodule:: ml4chem.data.handler
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+ml4chem.data.parser module
+--------------------------
+
+.. automodule:: ml4chem.data.parser
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.data.preprocessing module
 ---------------------------------
 
 .. automodule:: ml4chem.data.preprocessing
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.data.serialization module
 ---------------------------------
 
 .. automodule:: ml4chem.data.serialization
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+ml4chem.data.utils module
+-------------------------
+
+.. automodule:: ml4chem.data.utils
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.data.visualization module
 ---------------------------------
 
 .. automodule:: ml4chem.data.visualization
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 
 Module contents
 ---------------
 
 .. automodule:: ml4chem.data
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/source/ml4chem.fingerprints.rst b/docs/source/ml4chem.fingerprints.rst
@@ -8,39 +8,39 @@ ml4chem.fingerprints.autoencoders module
 ----------------------------------------
 
 .. automodule:: ml4chem.fingerprints.autoencoders
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.fingerprints.cartesian module
 -------------------------------------
 
 .. automodule:: ml4chem.fingerprints.cartesian
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.fingerprints.cutoff module
 ----------------------------------
 
 .. automodule:: ml4chem.fingerprints.cutoff
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.fingerprints.gaussian module
 ------------------------------------
 
 .. automodule:: ml4chem.fingerprints.gaussian
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 
 Module contents
 ---------------
 
 .. automodule:: ml4chem.fingerprints
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/source/ml4chem.models.rst b/docs/source/ml4chem.models.rst
@@ -8,55 +8,63 @@ ml4chem.models.autoencoders module
 ----------------------------------
 
 .. automodule:: ml4chem.models.autoencoders
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
+ml4chem.models.gaussian\_process module
+---------------------------------------
+
+.. automodule:: ml4chem.models.gaussian_process
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.models.kernelridge module
 ---------------------------------
 
 .. automodule:: ml4chem.models.kernelridge
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.models.loss module
 --------------------------
 
 .. automodule:: ml4chem.models.loss
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.models.merger module
 ----------------------------
 
 .. automodule:: ml4chem.models.merger
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.models.neuralnetwork module
 -----------------------------------
 
 .. automodule:: ml4chem.models.neuralnetwork
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.models.se3net module
 ----------------------------
 
 .. automodule:: ml4chem.models.se3net
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 
 Module contents
 ---------------
 
 .. automodule:: ml4chem.models
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/source/ml4chem.rst b/docs/source/ml4chem.rst
@@ -6,35 +6,44 @@ Subpackages
 
 .. toctree::
 
-    ml4chem.backends
-    ml4chem.data
-    ml4chem.fingerprints
-    ml4chem.models
+   ml4chem.backends
+   ml4chem.data
+   ml4chem.fingerprints
+   ml4chem.models
+   ml4chem.optim
 
 Submodules
 ----------
 
+ml4chem.metrics module
+----------------------
+
+.. automodule:: ml4chem.metrics
+   :members:
+   :undoc-members:
+   :show-inheritance:
+
 ml4chem.potentials module
 -------------------------
 
 .. automodule:: ml4chem.potentials
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 ml4chem.utils module
 --------------------
 
 .. automodule:: ml4chem.utils
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
 
 
 Module contents
 ---------------
 
 .. automodule:: ml4chem
-    :members:
-    :undoc-members:
-    :show-inheritance:
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/source/models.rst b/docs/source/models.rst
@@ -1,31 +1,47 @@
 ===================
-Atom-centered models
+Introduction
 ===================
 
 .. contents:: :local:
 
-Introduction
-============
 
-The `atomistc` module is designed to deploy atom-centered models with ml4chem.
+The `atomistic` module is designed to deploy models where the atom is the
+central object. Machine learning potentials might be one of the best known
+cases of atom-centered models. These models are known to be very accurate
+to predict energy and atomic forces and are very powerful because
+models can generalize to any molecular size as long as the training data
+contained atoms with similar chemical environments.
 
+========
 Theory
--------------------------------
+========
+The basic idea behind atomistic machine learning is that the prediction of a
+molecule or bulk can be obtained as the sum of atomic contributions:
 
-Some theory here.
+.. math::
 
+   P = \sum_{i=1}^n P_{atom}(R^{local})
 
-Fingerprints (atomic features)
+where :math:`P_{atom}` is a functional of atomic positions.
+
+==========================
+Atomic Features
 ==========================
 
-  - Binary (Pass/Fail)
-  - Multi (Cats, Dogs, Sheep)
-  - Ordinal (Low, Medium, High)
+ML4Chem support Gaussian symmetry functions and atomic latent features
+obtained with the `Autoencoder` class.
 
+  - Gaussian symmetry functions.
+  - Atomic latent features.
 
+
+==========================
 Models
 ==========================
 
+Deep Learning
+===============
+
 Neural Networks
 ----------------
 
@@ -37,7 +53,22 @@ Autoencoders
 
 Something here
 
+Support Vector Machines
+========================
+
 Kernel Ridge Regression
 ------------------------
 
 Something here. 
+
+Gaussian Process Regression
+------------------------
+
+Something here.
+
+===================
+Semi-supervised Learning
+===================
+
+Model Merger class
+============