Improved documentation for example scripts.

examples/README.md

@@ -1,7 +1,5 @@
-This directory contains a number of example experiments, scripts and jupyter notebooks.
+# Examples
 
-For all these examples it is assumed that the dependencies in environment.yml file have been installed.
+This directory contains legacy examples.
 
-If you are running locally the easiest way to install the dependencies is by using conda to create environment.yml.
-
-If you are viewing this on the web you use this [link](https://mybinder.org/v2/gh/VowpalWabbit/Coba/HEAD?filepath=examples/notebooks) to start a Jupyter environment with all dependencies installed.
+Many more examples can now be found in the [documentation](https://coba-docs.readthedocs.io/).

examples/scripts/Custom Vowpal.py

@@ -1,9 +1,8 @@
 """
-This is an example script that creates and executes an Experiment.
-This script requires that the matplotlib and vowpalwabbit packages be installed.
+This script creates and executes an Experiment with a custom namespace.
+This script requires the matplotlib and vowpalwabbit package.
 """
 
-import math
 import coba as cb
 
 class MakeCustomNamespaces:
@@ -17,6 +16,8 @@ def params(self):
         return {"target": self.target, "parts": self.partitions}
 
     def filter(self, interactions):
+        import math
+
         n_parts = len(self.partitions)
 
         for old in interactions:

examples/scripts/Getting Started.py

@@ -1,6 +1,6 @@
 """
-This is an example script that creates and executes an Experiment.
-This script depends on the matplotlib and vowpalwabbit packages.
+This script creates and executes an Experiment.
+This script requires the matplotlib and vowpalwabbit package.
 """
 
 import coba as cb

examples/scripts/Logged Results.py

@@ -1,3 +1,8 @@
+"""
+This script creates and executes an Experiment to generate logged data and then performs off-policy evaluation.
+This script requires the matplotlib and vowpalwabbit package.
+"""
+
 import coba as cb
 
 def main():
@@ -23,6 +28,7 @@ def main():
     log = cb.Environments.from_result(filename)
     lrn = cb.VowpalLearner("--cb_explore_adf --explore_eval")
     val = cb.SequentialCB(learn='off',eval=False)
+
     cb.Experiment(log, lrn, val).run()
 
     # performance in the original online run

examples/templates/README.md

@@ -1,5 +1,7 @@
+# Templates
+
 This directory contains two examples of environment template files.
 
 Environment templates are a great way to share experiments with other researchers.
 
-These files can be loaded via Environments.from_template("short_experiment.json").
+Template files can be loaded via a call such as Environments.from_template("short.json").