
 DONE            JOB
------------------------------------------------------------
  X     0. random seeding 
        1. warm-up : multiple starts
        2. autmatic extraction of features after learning
        3. automatic ranking on basis vectors after learning
  X     4. group-sparsity for multi-channel
  X     5. an script that generate and check config file  -------> a template is stored in the data folder
  X     6. multi-class/ multi-label case ----> for this initialization.m need to change
  X     7. writing MOSEK find module
        8. replacin mosek with a free version
        9. parallelize the projection algorithm
   X    10. ITK image read/write 
        11. In documentaion, talk about normalization
  N/A   12. Try/replace your spg with Mark Schmidt version (it should be very easy)
        13. Possibly replacing .mat with .hdf5
        14. Multi-resolution initialization for full resolution case
   X    15. Create a log file for each experiment
        16. Either change the way the program extract features or make temporary file inside of the experimehnt folder
        17. If you don't use className, remove it!
   X    18. When code continues, Reports is replaced with a new one fix it!
        19. Interface with Nipype: http://nipy.sourceforge.net/nipype/devel/matlab_interface_devel.html
        20. Add a sanity-check to paramSearch that each class should have more than one subject in order for param search to work
 
