README on how to make ECM2 tables.


1. make these directories, /training_data, /single_tables, /combined_tables, /your_archive

2. stick training_data into /training_data.   
   - copy or link /apollo/cloud/personal/heidinger/ecm2_training

3. If you have a new training data, use seperate_training_data_by_doy.pro to make test and train subsets.
    - use this naming convention: ecm_training_'+sensor+'.sav', ecm_training_'+sensor+'_test.sav', ecm_training_'+sensor+'_train.sav'
    -if you have already done this, skip this step
    - you want a ratio of 75% train and 25% test

3. Make the single classifier tables (1d, 2d, 3d).  
    -make an idl bat script for this and it takes awhile
    -don't use channels that don't exist on your sensor
    -this can take awhile
    -use "make_abhi_tables.bat" as reference.

4. Stick these in /single_tables  and archive them somewhere else


5. Use combine_single_tables.pro, this will combine the single tables in /combined_tables/test.nc

6. run test_on_training as follows:
     .rnew test_on_training
     test_on_training, sensor='xxxx', /optimize

     This will make figures and tables which are useful
     optimize option will optimize which classifiers are on or off
     
     it will also remove and alter tables in /single_tables

7. make a directory to store your results (/your_archive).
     -move /combined_tables/test.nc into there and call original_lut.nc
     -move optimal_output.txt in there
     -move figures in there

8. rerun combine_single_tables 

9. run test_on_training, sensor='xxxx', /add_thresh
    - this will make the surface dependent prob/conf numbers

10. move the /combined_tables/test.nc into your archive spot and rename like optimal_lut.nc

11. you are done