Write a class diagram for Shogun's GP framework

[karlnapf]

This is a preliminary task for the http://www.shogun-toolbox.org/page/Events/gsoc2014_ideas#variational_learning project

Since we will be adding some inference methods to the existing GP infrastructure in Shogun, it is essentially to understand how all the classes and methods play together. This task is to write a class diagram (in dia, like the one here https://github.com/karlnapf/shogun-mcmc-prototype/tree/master/doc/uml ) that tries to capture the general structure of the framework. This should include all important interfaces/methods/variables to illustrate the way we represent GP inference in Shogun. Try to avoid unnecessary details such as methods inherited from CSGObject - the diagram should be simple yet illustrative.

Eventually, we will be discussing how to implement the new algorithms within this diagram when the project is running.

Have a look at the unit tests and examples how methods work together

@votjakovr and @karlnapf are ready to take any questions on this - dont be afraid to ask :)

[abinashpanda]

I am attempting this issue :) . I have a doubt that the class diagram ( https://github.com/karlnapf/shogun-mcmc-prototype/tree/master/doc/uml) you have mentioned, whether it was generated from a script or manually created.

[karlnapf]

This was written manually. We could obviously create class diagrams from code automagically, but that misses the whole point of this task since computers are bad in distinguishing unimportant things from important ones :) This task is more about getting familiar with the framework to be ready to extend it. Also see the latest put up entrance tasks on variational learning #1971. This diagram might be useful for extending the framework for such tasks

[yorkerlin]

@karlnapf This is related to deep GPs.

[karlnapf]

yep