Source Code

• MLEngine  
  This takes a job file and uses it to:
  • Create a neural network as specified by the job file
  • Run the neural network for both training and testing data
  • Record the results in the job object
  In a sense, this is the interpreter for the job object. It is the actual work of running a neural network.
  
  
• MLJob  
  The is the runtime to read and write MLJob files, which are XML files used to configure a neural network as well as save and restore the state of a neural net. It is used for several applications:
  • Specify the training and testing of a neural net, and save the final matrices
  • Create a pre-trained neural net to make predictions
  • Pass network state between processes, either two processes on a single machine or else over a network from a controller machine to worker compute servers.
  
  
  
• TDF Tools  
  This implements TDF - Timeline Data Format. TDF is an xml file format that is designed to store time-series of data for medical applications.

  To read a TDF file, we typically iterate at several levels:

     For each patient in the file
         For each data entry in the current patient
  

  Additionally, you can identify several data windows within a single patient chart. An example data window may be all events between admission and discharge for a single hospital visit. So, you would instead iterate over:

         For each patient in the partition
             For each event window in the patient
                 For each data entry in the current window
  

• xmlTools  
  These are tools for manipulating an XML file, built on top of the minidom layer.
  
  
• testUtils.py   testUtils.py  
  Some utility procedures for testing and debugging.