*******************************************************************************************************
DBNizer
-------------------
-------------------

Hearm_Release_Jan_2015_baseline/

Structizer
------------

-Go to structuremake folder
-run networkx2.py 
   -make sure the input directed graph of the original system is in  'Structure58.txt'
   -Structure_init.txt should consist of the subset of nodes considered for the DBN (each variable in a newline)
   -output is a file 'Structure_Graph.txt' which are all the potential edges possible
   -copy this file into “data_files” folder
-compile and run MI_stats_var_conditional.cpp
    
takes as inputs :
"Structure_Graph.txt" in data_files/
"var_adjusted.csv" in data_files/
Simulation traces in  data_files/Simul_dump/

outputs: (currently the default is set to 5 parents limit, can be suitably changed )
the conditional MIs of all the different links (conditioned upto 3 variables)
"Structure.txt" in data_files/

CPTizer
-------

-DBN_make-fromtraces.cpp(build the CPTs)

input files:

"Structure.txt" in data_files/
"var_adjusted.csv" in data_files/
Simulation traces in  data_files/Simul_dump/

output:

"cpt.csv" (conditional probability tables for te structure)
"init_Marginal.csv" (initial marginal distribution)

additional files generated for debugging:
"cptdump.txt"
"cptdump_counts.txt"


Inference
---------------------
-DBN_inf_Hybrid_Simple.cpp (DBN inference on Nominal initial distribution)

input files:

"Structure.txt" in data_files/
"var_adjusted.csv" in data_files/
"cpt.csv" (conditional probability tables for te structure)
"init_Marginal.csv" (initial marginal distribution)

output:

"deat1.txt" (for every dead cell displays wen te cell died)
DBN simulation outputs in data_files/Simuldbn_dump/
"Marginals_new_Simul.txt" (Marginals of simulations)
The dead cells , alive cells and time of simulation are displayyed on page


-DBN_inf_Hybrid_Simple_makelist_partial.cpp (DBN inference on te 60 top configurations)

input files:

same as before,
additionally the set of configurations need be in data_files/init_configurations_reduced.txt, init_configurations_reduced-2.txt and init_configurations_reduced-1.txt

output files:
- init_death_time_distribtion_partial-2.txt,init_death_time_distribtion_partial-1.txt and init_death_time_distribtion_partial.txt
(Te contain the dead and alive statistics for each of the configurations)


Weighted error computation:
--------------------------
- error_DBN_HSD.py
computes the weighted error over all the 60 configurations reported in the paper 
reads the output files of previous step and death_init_HSD.txt (which contains the statistics of the HSD model)

output: produced a single file containing the variables of the DBN and contains the value of error in it

*****************************************************************************************************************

Single clean Automated pipeline (bash script):
-------------------

“script-singleclean” 

(runs the full pipeline described above) (invoke in terminal using ./script-singleclean)

**********************

Automated-iterative improvement (bash script)
——————————————

"script"

make sure the correct input file is read in tis base script i.e for instance if you already have a 7 variable DBN and want to evaluate by adding 1 more variable (then look for "Structure_init_7.txt" which has the 7 variables of the system, make sure to cange it for every subsequent iteration)
Structure58.txt contains all the variables of the system.

No other change is required.

The script automatically evaluates all the possible one variable additions and computes their error in datafile.