README

Description:
------------

  A small machine learning framework that does the following:
    - training data set generation
    - machine-learning model creation and saving to disk
    - invoking the model with test data instance(s)

  The framework has four executables (bash and python):
    - training_data_gen
      This is a shell script that takes input:
        -> a list of events to be measured; values of these
           events form a row of training data
           For a full list of supported events, execute:
             likwid-perfctr -e | grep ", PMC" | awk -F ',' '{print $1}'
        -> a list of programs and optional program arguments
           that will be profiled

      It outputs a text file consisting of
        -> a header line for the all the performance events
        -> as many rows as many programs, where each row
          contains the values of events that have been measured

    - train_ml
      This is a python script that takes input:
        -> a training data set, output by training_data_gen
        -> a target data set; this will need to be created 
           by the user depending on his/her objective
           function
        -> the name of the binary file where the trained
           model will be written to
    
      It outputs a binary file that contains the serialized
      model object

    - test_ml
      This is a python script that takes input:
        -> a test data set, formatted similarly to the
           training data set
        -> a binary file containing the model to be
           invoked

      It outputs the results from the model
  
    - perf_counter
      This is a shell script that is called internally to measure
      performance events. This is the file that should be 
      modified if the user wants to add code to automatically
      generate target data based on his/her objective function

    - get_hex_codes_from_names
      This is a shell script that is called internally to
      convert the provided names of the performance events to
      hex codes

Dependencies:
-------------

  - likwid
  - bash shell
  - python
    - scikit-learn
    - numpy
  
Installation:
-------------

  After checking out the SVN repository,
  add the bin directory to your path