==========================================================================================
README for reproducibility submission of paper ID 940

A) Source code info

Repository: https://github.com/dongyoungy/dbsherlock-reproducibility
Programming Language: MATLAB
Additional Programming Language info: N/A
Compiler Info: N/A
Packages/Libraries Needed: MATLAB R2015b or higher

B) Datasets info

Repository: https://github.com/dongyoungy/dbsherlock-reproducibility/tree/master/datasets
Data generators: N/A (datasets are provided in the repository)

C) Hardware Info

The reproducibility package has been tested with my mid 2014 Macbook Pro Retina with
the following specification:
C1) Processor: 2.5Ghz quad-core Intel Core i7 processor (Turbo Boost up to 3.7Ghz)
C2) Caches: 1MB L2 cache, 6MB shared L3 cache
C3) Memory: 16GB 1600Mhz DDR3L SDRAM
C4) Secondary Storage: 512 GB PCIe-based flash storage
C5) Network: N/A

D) Experimentation Info

D1) Dataset archives that were used to conduct experiments in the paper have been
provided in the repository. Each dataset is packaged as .mat file in ./datasets directory.
The directory has a separate README file, which explains the format of the datasets for
people who want to explore and use datasets only.
D2) In order to run experiments, launch MATLAB and go to the root directory of the
repository using the command window. Once you enter the command 'reproduce_cui' to run our
script in the command window, You will see the following menu, where you can select an
experiment that you want to reproduce from the paper:

                << DBSherlock Experiments >>
                1. Accuracy of Single Causal Models (Section 8.3)
                2. DBSherlock Predicates versus PerfXplain (Section 8.4)
                3. Effectiveness of Merged Causal Models (Section 8.5)
                4. Effect of Incorporating Domain Knowledge (Section 8.6)
                5. Explaining Compound Situations (Section 8.7)
                6. Run all of the above (ETC: 4-5 hours)

      Select an experiment to reproduce (1-6 or other input to exit):

You can simply choose the option '6' to reproduce all experiments. The estimated time of
completion has been obtained by running it with my test machine. Since it is mostly
CPU-bound, you will probably get the result faster with modern desktop machines. Once it
is done, it will plot the charts that replicate figures and tables in the experiment
section of the paper automatically (except for the section '8.8 User Study').

==========================================================================================