switch to aesara; formatting

Author: jkaardal

README.md

@@ -12,17 +12,15 @@ inequality constraints, {ci}. PYIPM uses a line search interior-point
 method to search for an x that is a feasible local minimizer of a
 user-defined nonlinear program.
 
-## Shift to Theano-PyMC
+## Shift to Aesara
 
 Originally, this module used the Theano library maintained by the
 Montreal Institute for Learning Algorithms. The original version of
-Theano was deprecated in 2017. Fortunately, the PyMC team has taken
+Theano was deprecated in 2017. Fortunately, the Aesara dev team has taken
 over maintenance of Theano in a fork called
-[Theano-PyMC](https://github.com/pymc-devs/Theano-PyMC). The install
-script (setup.py) now installs Theano-PyMC instead of Theano as a
-dependency. That said, this module is still compatible with deprecated
-Theano so you may choose to continue using deprecated Theano (at least
-as long as your architecture can still run it).
+[Aesara](https://github.com/aesara-devs/aesara). The install
+script (setup.py) now installs Aesara instead of Theano as a
+dependency.
 
 ## Getting Started
 
@@ -35,14 +33,14 @@ Using PYIPM requires either a python 3 or python 2.7 interpreter. In
 addition to having python installed, the following packages are
 required:
 
-    NumPy, SciPy, Theano-PyMC
+    NumPy, SciPy, Aesara
 
-Theano is used to speed up the code execution through parallelization
+Aesara is used to speed up the code execution through parallelization
 and compilations of certain functions used in PYIPM. Furthermore,
-Theano makes it possible to perform many of largest steps in the
+Aesara makes it possible to perform many of largest steps in the
 algorithm on graphics processing units (GPUs). For information on
-using the GPUs, see the Theano documentation at
-http://deeplearning.net/software/theano/.
+using the GPUs, see the Aesara documentation at
+https://aesara.readthedocs.io/en/latest/.
 
 All of the above required packages may be installed through pip.
 
@@ -128,14 +126,14 @@ minimum if you compare the 'Ground truth' line to the 'Solver
 solution' line.
 
 These example problems can also be used as a guide to construct your
-own problems which may be helpful if you are unfamiliar with Theano.
+own problems which may be helpful if you are unfamiliar with Aesara.
 
 ## Building and solving your own problems
 
 The basics of writing your own problem statements are to define the
 objective function, f, the equality constraints function, ce, and the
 inequality constraints function, ci, as symbolic expressions that
-depend on the Theano tensor variable, x_dev. The tensor variable x_dev
+depend on the Aesara tensor variable, x_dev. The tensor variable x_dev
 is the symbolic input to all of the aforementioned functions that
 represents the weights, x.
 
@@ -175,17 +173,17 @@ kkt = the Karush-Kuhn-Tucker (KKT) conditions (length 4 list of NumPy
     arrays)
 
 The reason why compiling f, ce, and ci would cause errors in this case
-is because PYIPM uses Theano's symbolic differentation functionality
+is because PYIPM uses Aesara's symbolic differentation functionality
 to generate expressions for the gradients, df, dce, and dci, and, if
 the exact Hessian is being used, the Hessians, d2f, d2ce, and
 d2ci. Unfortunately, once these expressions are compiled, they cannot
 be used for symbolic differentiation. That is not to say, however,
-that you cannot use theano.function() to precompile your expressions
+that you cannot use aesara.function() to precompile your expressions
 into functions. You can do that at your own convenience so long as you
 provide your own symbolic expressions or functions for df, dce, and
 dci and, if applicable, d2f, d2ce, and d2ci. It is worth pointing out,
 however, that PYIPM does much more than just automatic differentiation
-(through Theano) and compiling your expressions; PYIPM also organizes
+(through Aesara) and compiling your expressions; PYIPM also organizes
 these inputs into an overall gradient and Hessian. To take full
 advantage of PYIPM for larger problems, it is best to provide PYIPM
 with expressions such that it can construct and compile the symbolic
@@ -218,8 +216,8 @@ it to 0 or False.
 
 For a more thorough guide to the parameters and public class
 functions, see the docstring of class IPM in pyipm.py. For more
-information about how to build and compile expressions, see the Theano
-documentation at http://deeplearning.net/software/theano/.
+information about how to build and compile expressions, see the Aesara
+documentation at https://aesara.readthedocs.io/en/latest/.
 
 # Contributing, help, bug reporting, etc.