reorganize inference docs api

docs/autogen.py

@@ -64,7 +64,6 @@
             'inference-compositionality.tex',
             'inference-data-subsampling.tex',
             'inference-development.tex',
-            'inference-api.tex',
         ],
     },
     {
@@ -99,14 +98,6 @@
         ],
         'child_pages': [],
     },
-    {
-        'page': 'inference-api.tex',
-        'title': 'API',
-        'parent_pages': [
-            'inference.tex'
-        ],
-        'child_pages': [],
-    },
     {
         'page': 'criticism.tex',
         'title': 'Criticism',

docs/tex/api/criticism.tex

@@ -35,6 +35,8 @@ \subsubsection{Criticism}
 % Developing new criticism techniques is easy.  They can be derived from
 % the current techniques or built as a standalone function.
 
+\begin{center}\rule{3in}{0.4pt}\end{center}
+
 {{sphinx
 
 .. automodule:: edward.criticisms

docs/tex/api/inference-classes.tex

@@ -103,4 +103,43 @@ \subsubsection{Exact Inference}
 classical Gibbs and mean-field updates \citep{bishop2006pattern} without
 tedious algebraic manipulation.
 
+\begin{center}\rule{3in}{0.4pt}\end{center}
+
+{{sphinx
+
+.. autoclass:: edward.inferences.KLqp
+   :members:
+
+.. autoclass:: edward.inferences.ReparameterizationKLqp
+
+.. autoclass:: edward.inferences.ReparameterizationKLKLqp
+
+.. autoclass:: edward.inferences.ReparameterizationEntropyKLqp
+
+.. autoclass:: edward.inferences.ScoreKLqp
+
+.. autoclass:: edward.inferences.ScoreKLKLqp
+
+.. autoclass:: edward.inferences.ScoreEntropyKLqp
+
+.. autoclass:: edward.inferences.KLpq
+   :members:
+
+.. autoclass:: edward.inferences.MAP
+   :members:
+
+.. autoclass:: edward.inferences.Laplace
+   :members:
+
+.. autoclass:: edward.inferences.MetropolisHastings
+   :members:
+
+.. autoclass:: edward.inferences.HMC
+   :members:
+
+.. autoclass:: edward.inferences.SGLD
+   :members:
+
+}}
+
 \subsubsection{References}\label{references}

docs/tex/api/inference-development.tex

@@ -8,62 +8,42 @@ \subsubsection{Developing Inference Algorithms}
 methods. This enables fast experimentation on top of existing
 algorithms, whether it be developing new black box algorithms or
 new model-specific algorithms.
+For examples of algorithms developed in Edward, see the inference
+\href{/tutorials/}{tutorials}.
 
 \includegraphics[width=700px]{/images/inference_structure.png}
 {\small\textit{Dependency graph of inference methods.
 Nodes are classes in Edward and arrows represent class inheritance.}}
 
 There is a base class \texttt{Inference}, from which all inference
-methods are derived from.
-
-\begin{lstlisting}[language=Python]
-class Inference(object):
-  """Base class for Edward inference methods.
-  """
-  def __init__(self, latent_vars=None, data=None, model_wrapper=None):
-    ...
-\end{lstlisting}
-
-It takes as input the set of latent variables to infer and a dataset. Optionally, if the user uses an external language to specify the model, it takes as input a model wrapper \texttt{model_wrapper}.
-
-Note that \texttt{Inference} says nothing about the class of models that an
-algorithm must work with. One can build inference algorithms which are
-tailored to a restricted class of models available in Edward (such as
-differentiable models or conditionally conjugate models), or even
-tailor it to a single model. The algorithm can raise an error if the
-model is outside this class.
+methods are derived from. Note that \texttt{Inference} says nothing
+about the class of models that an algorithm must work with. One can
+build inference algorithms which are tailored to a restricted class of
+models available in Edward (such as differentiable models or
+conditionally conjugate models), or even tailor it to a single model.
+The algorithm can raise an error if the model is outside this class.
 
 We organize inference under two paradigms:
 \texttt{VariationalInference} and \texttt{MonteCarlo} (or more plainly,
 optimization and sampling). These inherit from \texttt{Inference} and each
 have their own default methods.
 
-\begin{lstlisting}[language=Python]
-class MonteCarlo(Inference):
-  """Base class for Monte Carlo inference methods.
-  """
-  def __init__(latent_vars, data=None, model_wrapper=None):
-    super(MonteCarlo, self).__init__(latent_vars, data, model_wrapper)
+For example, developing a new variational inference algorithm is as simple as
+inheriting from \texttt{VariationalInference} or one of its derived
+classes. \texttt{VariationalInference} implements many default methods such
+as \texttt{initialize()} with options for an optimizer.
 
-  ...
+\begin{center}\rule{3in}{0.4pt}\end{center}
 
+{{sphinx
 
-class VariationalInference(Inference):
-  """Base class for variational inference methods.
-  """
-  def __init__(self, latent_vars=None, data=None, model_wrapper=None):
-    super(VariationalInference, self).__init__(latent_vars, data, model_wrapper)
+.. autoclass:: edward.inferences.Inference
+   :members:
 
-  ...
-\end{lstlisting}
+.. autoclass:: edward.inferences.VariationalInference
+   :members:
 
-For example, developing a new variational inference algorithm is as simple as
-inheriting from \texttt{VariationalInference} or one of its derived
-classes. \texttt{VariationalInference} implements many default methods such
-as \texttt{initialize()} with options for an optimizer.
+.. autoclass:: edward.inferences.MonteCarlo
+   :members:
 
-For examples of inference algorithms developed in Edward, see the inference
-\href{/tutorials/}{tutorials}. It can also be useful to look at
-the
-\href{https://github.com/blei-lab/edward/tree/master/edward/inferences}
-{source code}.
+}}