Tidy up tutorial solution notebooks and provide problem templates about reinforcement_learning_course_materials

[wkirgsn]

Generation of "emptied" solution files that act as templates is perfectly possible with nbgrader, just enclosing stuff that needs to be hidden with "### BEGIN SOLUTION" and "# END SOLUTION".

Not sure what happens with travis-generated content, though. I think it can be commited next to the solution files with a small script in travis, then we would have a travis commit after each of our commits

[Webbah]

Does the enclosing works also for markdown?
If not, how do we mark solutions there?

[wkirgsn]

Ok this is the instruction list:

1. Install nbgrader with

conda install jupyter
conda install -c conda-forge nbgrader

(do not use pip!)

2. Get the folder structure right, such that nbgrader can work:

   • exercises/
     • nbgrader_config.py # configuration
     • templates/ # destination for autogenerated templates
     • solutions/ # actual source code (solutions with markers)
       • ex1/
       • ex2/
         ...

3. For each exercise solution notebook:

   • Open the notebook and click on View > Cell toolbar > Create Assignment
     • This enables assigning types and points to each cell
   • Mark task templates with "read-only", solution markdown with "manually graded answer", and solution code with "autograded answer", where solution lines are enclosed with ### BEGIN SOLUTION and ### END SOLUTION.

4. Generate the template from the solution with nbgrader generate_assignment "<<exercise name>>",
   e.g. nbgrader generate_assignment "ex01"

I can go ahead and start with point 2 but I will need assistance for point 3 and all exercises except 1 and 8.
The generation of the templates (point 4) from the solutions can be outsourced to TravisCI as soon as we have 1-3 finished.

[wkirgsn]

all templates are generated and readmes were created.