Future of Shogun Brainstorming
Soeren Sonnenburg edited this page Jul 12, 2013
·
8 revisions
- OpenMP support
- replace SG_REF + SG_UNREF with smart pointers (maybe c++11?)
- support for drop-in replacement for std malloc (memory managment). e.g. jemalloc
- thread pool
- add CSGObject::clone() that implements clone() for all the derived classes
- implement baseline ML algorithms that are almost in every other ML library are available (random forests, adaboost)
- documention/howto/examples for each and every algorithm we have in shogun. the examples preferable implemented in python
- large scale
- SGString -> SGReferencedData
- nice model selection syntax for convenient model selection
- 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