Get the transform docs into the API pages.

RELEASE-NOTES.md

@@ -32,6 +32,7 @@
 - Wrapped `DensityDist.rand` with `generate_samples` to make it aware of the distribution's shape. Added control flow attributes to still be able to behave as in earlier versions, and to control how to interpret the `size` parameter in the `random` callable signature. Fixes [3553](https://github.com/pymc-devs/pymc3/issues/3553)
 - Added `theano.gof.graph.Constant` to type checks done in `_draw_value` (fixes issue [3595](https://github.com/pymc-devs/pymc3/issues/3595))
 - `HalfNormal` did not used to work properly in `draw_values`, `sample_prior_predictive`, or `sample_posterior_predictive` (fixes issue [3686](https://github.com/pymc-devs/pymc3/pull/3686))
+- Random variable transforms were inadvertently left out of the API documentation. Added them. (See PR [3690](https://github.com/pymc-devs/pymc3/pull/3690)).
 
 ## PyMC3 3.7 (May 29 2019)
 

docs/source/api.rst

@@ -20,6 +20,7 @@ API Reference
    api/data
    api/model
    api/model_graph
+   api/variables
    api/shape_utils
 
 

docs/source/api/distributions.rst

@@ -9,4 +9,5 @@ Distributions
    distributions/multivariate
    distributions/mixture
    distributions/timeseries
+   distributions/transforms
    distributions/utilities

docs/source/api/distributions/transforms.rst

@@ -0,0 +1,143 @@
+****************************************************************
+Transformations of a random variable from one space to another.
+****************************************************************
+
+Note that for convenience these entities can be addressed as
+``pm.transforms.``\ *X* for any name *X*, although they are actually
+implemented as ``pm.distributions.transforms.``\*X*.
+
+.. currentmodule:: pymc3.distributions.transforms
+
+
+.. contents ::
+
+..
+   .. autosummary::
+
+   transform
+   stick_breaking
+   logodds
+   interval
+   log_exp_m1
+   lowerbound
+   upperbound
+   ordered
+   log
+   sum_to_1
+   t_stick_breaking
+   circular
+   CholeskyCovPacked
+   Chain
+
+
+
+Transform Instances
+~~~~~~~~~~~~~~~~~~~
+
+Transform instances are the entities that should be used in the
+``transform`` parameter to a random variable constructor.  These are
+initialized instances of the Transform Classes, which are described
+below.
+
+.. glossary::
+``stick_breaking``
+    Instantiation of :class:`~pymc3.distributions.transforms.StickBreaking`
+    :class:`~pymc3.distributions.transforms.Transform` class for use in the ``transform``
+    argument of a random variable.
+
+``logodds``
+    Instantiation of
+    :class:`~pymc3.distributions.transforms.LogOdds` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+``interval``
+    Alias of
+    :class:`~pymc3.distributions.transforms.Interval` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+``log_exp_m1``
+    Instantiation of
+    :class:`~pymc3.distributions.transforms.LogExpM1` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+``lowerbound``
+    Alias of
+    :class:`~pymc3.distributions.transforms.LowerBound` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+``upperbound``
+    Alias of
+    :class:`~pymc3.distributions.transforms.UpperBound` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+``ordered``
+    Instantiation of
+    :class:`~pymc3.distributions.transforms.Ordered` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+``log``
+    Instantiation of
+    :class:`~pymc3.distributions.transforms.Log` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+
+``sum_to_1``
+    Instantiation of
+    :class:`~pymc3.distributions.transforms.SumTo1` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+
+``circular``
+    Instantiation of
+    :class:`~pymc3.distributions.transforms.Circular` :class:`~pymc3.distributions.transforms.Transform` class
+    for use in the ``transform`` argument of a random variable.
+
+
+.. autofunction::  t_stick_breaking 
+
+
+Transform Base Classes
+~~~~~~~~~~~~~~~~~~~~~~
+
+Typically the programmer will not use these directly.
+
+.. autoclass::  Transform
+    :members:
+.. autoclass::  transform
+    :members:
+.. autoclass::  TransformedDistribution
+    :members:
+
+
+Transform Composition Classes
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass:: Chain
+    :members:
+.. autoclass:: CholeskyCovPacked
+    :members:
+
+
+Specific Transform Classes
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autoclass::  Log
+    :members:
+.. autoclass::  LogExpM1
+    :members:
+.. autoclass::  LogOdds
+    :members:
+.. autoclass::  Interval
+    :members:
+.. autoclass::  LowerBound
+    :members:
+.. autoclass::  UpperBound
+    :members:
+.. autoclass::  Ordered
+    :members:
+.. autoclass::  SumTo1
+    :members:
+.. autoclass::  StickBreaking
+    :members:
+.. autoclass::  Circular
+    :members:

docs/source/api/variables.rst

@@ -0,0 +1,28 @@
+Random Variables
+----------------
+
+.. currentmodule:: pymc3.model
+
+The normal PyMC3 programmer will typically not need to interact with these classes, except possibly when debugging.  Otherwise they are primarily of interest to developers.
+
+
+.. autoclass:: PyMC3Variable
+    :members:
+
+
+.. autoclass:: ValueGradFunction
+    :members:
+
+
+.. autoclass:: FreeRV
+    :members:
+
+.. autoclass:: ObservedRV
+    :members:
+
+.. autoclass:: MultiObservedRV
+    :members:
+
+.. autoclass:: TransformedRV
+    :members:
+

docs/source/developer_guide.rst

@@ -378,25 +378,29 @@ https://github.com/pymc-devs/pymc3/blob/6d07591962a6c135640a3c31903eba66b34e71d8
         self.add_random_variable(var)
         return var
 
-In general, if there is observed, the RV is defined as a ``ObservedRV``,
-otherwise if it has a transformed method, it is a ``TransformedRV``, otherwise, it returns the
-most elementary form: a ``FreeRV``.
+In general, if a variable has observations (``observed`` parameter), the RV is defined as an ``ObservedRV``,
+otherwise if it has a ``transformed`` (``transform`` parameter) attribute, it is a
+``TransformedRV``, otherwise, it will be the most elementary form: a
+``FreeRV``.  Note that this means that random variables with
+observations cannot be transformed.
 
-Below, I will take a deeper look into ``TransformedRV``, a normal user
+Below, I will take a deeper look into ``TransformedRV``. A normal user
 might not necessary come in contact with the concept, as
 ``TransformedRV`` and ``TransformedDistribution`` are intentionally not
 user facing.
 
 Because in PyMC3 there is no bijector class like in TFP or pyro, we only
 have a partial implementation called ``Transform``, which implements
 Jacobian correction for forward mapping only (there is no Jacobian
-correction for inverse mapping). The use case we considered are limited
+correction for inverse mapping). The use cases we considered are limited
 to the set of distributions that are bounded, and the transformation
 maps the bounded set to the real line - see
-`doc <https://docs.pymc.io/notebooks/api_quickstart.html#Automatic-transforms-of-bounded-RVs>`__.
+`doc
+<https://docs.pymc.io/notebooks/api_quickstart.html#Automatic-transforms-of-bounded-RVs>`__.
+However, other transformations are possible.
 In general, PyMC3 does not provide explicit functionality to transform
 one distribution to another. Instead, a dedicated distribution is
-usually created in consideration of optimising performance. But getting a
+usually created in order to optimise performance. But getting a
 ``TransformedDistribution`` is also possible (see also in
 `doc <https://docs.pymc.io/notebooks/api_quickstart.html#Transformed-distributions-and-changes-of-variables>`__):
 
@@ -422,7 +426,7 @@ usually created in consideration of optimising performance. But getting a
 
 
 
-Now, back to ``model.RV(...)`` - things return from ``model.RV(...)``
+Now, back to ``model.RV(...)`` - things returned from ``model.RV(...)``
 are Theano tensor variables, and it is clear from looking at
 ``TransformedRV``:
 
@@ -431,19 +435,19 @@ are Theano tensor variables, and it is clear from looking at
     class TransformedRV(TensorVariable):
         ...
 
-as for ``FreeRV`` and ``ObservedRV``, they are TensorVariable with
-Factor:
+as for ``FreeRV`` and ``ObservedRV``, they are ``TensorVariable``\s with
+``Factor`` as mixin:
 
 .. code:: python
 
     class FreeRV(Factor, TensorVariable):
         ...
 
-and ``Factor`` basically `enable and assign the
+``Factor`` basically `enable and assign the
 logp <https://github.com/pymc-devs/pymc3/blob/6d07591962a6c135640a3c31903eba66b34e71d8/pymc3/model.py#L195-L276>`__
 (representated as a tensor also) property to a Theano tensor (thus
-making it a random variable). For a ``TransformedRV``, it transform the
-distribution into a ``TransformedDistribution``, and then model.Var is
+making it a random variable). For a ``TransformedRV``, it transforms the
+distribution into a ``TransformedDistribution``, and then ``model.Var`` is
 called again to added the RV associated with the
 ``TransformedDistribution`` as a ``FreeRV``:
 
@@ -483,11 +487,11 @@ the model logp), and also deterministic transformation (as bookkeeping):
 named\_vars, free\_RVs, observed\_RVs, deterministics, potentials,
 missing\_values. The model context then computes some simple model
 properties, builds a bijection mapping that transforms between
-dictionary and numpy/Theano ndarray, thus allowing logp/dlogp function
-to have two equivalent version: one take a dict as input and the other
-take a ndarray as input. More importantly, a pm.Model() contains methods
-to compile Theano function that takes Random Variables (that are also
-initialised within the same model) as input.
+dictionary and numpy/Theano ndarray, thus allowing the ``logp``/``dlogp`` functions
+to have two equivalent versions: one takes a ``dict`` as input and the other
+takes an ``ndarray`` as input. More importantly, a ``pm.Model()`` contains methods
+to compile Theano functions that take Random Variables (that are also
+initialised within the same model) as input, for example:
 
 .. code:: python
 

pymc3/__init__.py

@@ -3,6 +3,7 @@
 
 from .blocking import *
 from .distributions import *
+from .distributions import transforms
 from .glm import *
 from . import gp
 from .math import logaddexp, logsumexp, logit, invlogit, expand_packed_triangular, probit, invprobit

pymc3/distributions/__init__.py

@@ -81,12 +81,6 @@
 from .timeseries import MvGaussianRandomWalk
 from .timeseries import MvStudentTRandomWalk
 
-from .transforms import transform
-from .transforms import stick_breaking
-from .transforms import logodds
-from .transforms import log
-from .transforms import sum_to_1
-
 from .bound import Bound
 
 __all__ = ['Uniform',