# CD appendix for Demet Ozgur Unluakin Ph.D thesis Files in the Disk : ------------------- Maintenance_under_partial_observations/experimental design/aggregation/aggregate_data_sets.zip : Matlab codes to aggregate data sets and to prepare finite-horizon ZMDP input files in pomdp-solve program format. Each aggregate_dxx.m produces two important output files: agg_agg_dxx.pomdp and zmdp_dxx.pomdp Maintenance_under_partial_observations/experimental design/aggregation/Aggregate_POMDP_files : Aggregated files (agg_agg_dxx.pomdp) of data sets d01-d14 wrt states and actions in pomdp-solve program format. Aggregated files of data sets d15-d28 are exactly the same as of d01-d14. Aggregated file agg_agg_dxx.pomdp is later used to exactly solve the aggregated POMDP problem of data set dxx with pomdp-solve program Maintenance_under_partial_observations/experimental design/aggregation/ZMDP_files : Finite-horizon ZMDP input files of data sets in pomdp-solve program format. If the file contains the initial belief (start value of pomdp-solve program format), it is named as zmdp_dxx.pomdp. Otherwise it is named as zmdp_dxx_ns.pomdp. zmdp_dxx.pomdp or zmdp_dxx_ns.pomdp is later used to approximately solve the original POMDP problem of data set dxx with ZMDP Maintenance_under_partial_observations/experimental design/disaggregation/disaggregate_data_sets.zip : Matlab codes and aggregate policy files (obtained after solving agg_agg_dxx.pomdp with pomdp-solve program) to disaggregate data sets Each disaggregate_dxx_new.m disaggregates its corresponding aggregate policy file. Maintenance_under_partial_observations/experimental design/disaggregation/solutions : Disaggregation results of data sets belonging to 50 replications. Cost, number of replacements and number of red observations of each replication are reported. Sample mean and standard deviation values are also reported. Maintenance_under_partial_observations/experimental design/re_aggregation/re_aggregate_data_sets.zip : Matlab codes to reaggregate data sets with weights Maintenance_under_partial_observations/experimental design/re_aggregation/Re_aggregate_POMDP_files : Weighted aggregated files (re_agg_agg_dxx.pomdp) of data sets Maintenance_under_partial_observations/experimental design/re_disaggregation/re_disaggregate_data_sets.zip : Matlab codes and reaggregate policy files (obtained after solving re_agg_agg_dxx.pomdp with pomdp-solve program) to redisaggregate data sets Each disaggregate_dxx_re.m redisaggregates its corresponding reaggregate policy file. Maintenance_under_partial_observations/experimental design/re_disaggregation/solutions : Disaggregation results of weighted aggregation of data sets belonging to 50 replications. Maintenance_under_partial_observations/experimental design/ZMDP_simulation/simulate_zmdp_data_sets.zip : Matlab codes and ZMDP policy files to simulate ZMDP policies of data sets. Each disaggregate_zmdp_dxx(_ns).m simulates its corresponding ZMDP policy file. Maintenance_under_partial_observations/experimental design/ZMDP_simulation/solutions : ZMDP simulation results of data sets belonging to 50 replications. Cost, number of replacements and number of red observations of each replication are reported. Sample mean and standard deviation values are also reported. Maintenance_under_partial_observations/k_out_of_n_systems/aggregation/aggregate_data_sets.zip : Matlab codes to aggregate data sets and to prepare finite-horizon ZMDP input files in pomdp-solve program format for k-out-of-4 systems. Each aggregate_dxx_4nyk.m produces two important output files: agg_agg_dxx_4nyk.pomdp and zmdp_dxx_4nyk.pomdp Maintenance_under_partial_observations/k_out_of_n_systems/aggregation/Aggregate_POMDP_files : Aggregated files (agg_agg_dxx_4nyk.pomdp) of data sets of k-out-of-4 systems wrt states and actions in pomdp-solve program format. Aggregated file agg_agg_dxx_4nyk.pomdp is later used to exactly solve the aggregated POMDP problem of data set dxx_4nyk with pomdp-solve program Maintenance_under_partial_observations/k_out_of_n_systems/aggregation/ZMDP_files : Finite-horizon ZMDP input files of data sets of k-out-of-4 systems in pomdp-solve program format. zmdp_dxx_4nyk.pomdp is later used to approximately solve the original POMDP problem of data set dxx_4nyk with ZMDP Maintenance_under_partial_observations/k_out_of_n_systems/disaggregation/disaggregate_data_sets.zip : Matlab codes and aggregate policy files (obtained after solving agg_agg_dxx_4nyk.pomdp with pomdp-solve program) to disaggregate data sets of for k-out-of-4 systems. Each disaggregate_dxx_4nyk.m disaggregates its corresponding aggregate policy file. Maintenance_under_partial_observations/k_out_of_n_systems/disaggregation/solutions : Disaggregation results of data sets of k_out_of_4_systems belonging to 50 replications. Cost, number of replacements and number of red observations of each replication are reported. Sample mean and standard deviation values are also reported. Maintenance_under_partial_observations/k_out_of_n_systems/ZMDP_simulation : Matlab codes and ZMDP policy (d14_4n4k.policy) file to simulate ZMDP policy of data set d14_4n4k. "disaggregate_zmdp_d14_4n4k.m" simulates its corresponding ZMDP policy file (d14_4n4k.policy). "zmdp_d14_4n4k.txt" is the solution file of ZMDP simulation of data set d14_4n4k. Maintenance_without_observations/DAP : Contains sources for DBN algorithm with FEM, FEL, REM, REL, FEMc, FELc, REMc, RELc and Random component selection approaches to solve DAP. Maintenance_without_observations/DAP/solutions : Sample solutions for DAP Maintenance_without_observations/Two_comp_sys_ex : Contains sources for DBN algorithm with FEM, FEL, FEnL and REM component selection approaches to solve the two component system example. Maintenance_without_observations/Two_comp_sys_ex/solutions : Sample solutions for the two component system example. ************************************************************************************************************************************************************************************ Hardware Requirements : ----------------------- * Intel Pentium based PC * Windows 2000 or Windows XP running hardware * 20 GB Hard Disk Space * 500 MB Main RAM ************************************************************************************************************************************************************************************ Software Requirements : ----------------------- * Windows 2000 or Windows XP * MATLAB 6.5 * BNT (Bayes Net Toolbox for MATLAB available from http://www.cs.ubc.ca/~murphyk/Software/BNT/bnt.html)