Future of Shogun Brainstorming

shogun features wishlist:

1. OpenMP support
2. replace SG_REF + SG_UNREF with smart pointers (maybe c++11?)
3. support for drop-in replacement for std malloc (memory managment). e.g. jemalloc
4. thread pool
5. add CSGObject::clone() that implements clone() for all the derived classes
6. implement baseline ML algorithms that are almost in every other ML library are available (random forests, adaboost)
7. documention/howto/examples for each and every algorithm we have in shogun. the examples preferable implemented in python
8. large scale
9. SGString -> SGReferencedData
10. nice model selection syntax for convenient model selection
11. cmake based configure

• swig R
• compile wrapper code size
• switch to cmake
• wish to install a binary for the user
• ipython notebook for documentation
• nightly binary builds .tar .rpm .deb