TODO.md

• Output normalization for individual forests
• Output normalization for ensembles
• Interface to Matlab for training
  • Matlab tree class files
  • Load data into class file
  • Background C++ process that receives firing rate from Matlab and returns class membership
    • Function to resample raw data onto normalized frame
    • Function to classify using interpolation
      • Append datafile properties to tree
• Improve commandline options
  • Allow to identify desired folds for transform
  • Ask user to confirm overwrites of files and allow -force option
  • Allow for unordered input
• Add support for continuation of interrupted training
  • Save RNG state
  • Parse trees for completeness
  • Add commandline options
• Reorganize outputs
  • Save ECOC matrix to separate file and allow matrix ID as optional input
  • Save training times in params directory with reference to ecoc matrix
  • Include comments in params file to clarify values?
  • Remove transform status
  • Save test results in more verbose files
• Generation of parameter files using Bayesian optimization
  • in Matlab
  • in C++
• Closed loop batching optimization process
• Fix node indices from double to unsigned int
• Cleanup code for consistent typing, iterands...
  • use size_t
  • const where appropriate