- improve command line 
  - run/build only one implementation/algorithm

- add more analysis
  - lines of code
  - density

- improve performance
  - memory: currently all the models are loaded in memory