Bao Thai Tran, Ngoc Hoang Luong
make
./run_experiments
After performing experiments of MEAs, the directory structure of results should be:
results
└─── 0_0
│ └─── p1
│ └─── run0
│ │ population00001_elites.dat
│ │ ...
│ ...
where,
0_0contains results obtained by using the algorithm0_0.p1is the problem 1 in the CEC2013 niching benchmark.run0is a run when solving the problem 1.population00001_elites.datcontains all potential solutions found by the algorithm in the iteration 1.
results contains only raw data which must be extracted to drawable data via src/mdstn/extract_data.py
There are two variables in this file extract_data.py which should be noticed such as algo and prob_ids:
algois a dictionary whose key is name of folder containing results of an algorithm and value is the algorithm's name.prob_idsis a list of indices of problems solved by the algorithm.
After making the necessary adjustment, the user takes command python src/mdstn/extract_data.py and obtains the cec2013 folder contains multiple files algo_prob_dims.txt.
- Convert
algo_prob_dims.txtto an appropriate extensionalgo_prob_dims.RDataby
Rscript src/mdstn/alg-create.R <path to folder containing .txt> <number of runs>
Example: Rscript src/mdstn/alg-create.R cec2013/p1 50
- Plot an MDSTN for each algorithm by
Rscript src/mdstn/alg-plot.R <path to folder containing .RData> <scale factor>
Example: Rscript src/mdstn/alg-plot.R cec2013/p1-stn 1.0
- Combine to a merged data file
.RDataby
Rscript src/mdstn/merged-create.R <path to folder containing .RData>
Example: Rscript src/mdstn/merged-create.R cec2013/p1-stn
- Plot a merging MDSTN for multiple algorithms by
Rscript src/mdstn/merged-plot.R <path to merged file .RData> <scale factor>
Example: Rscript src/mdstn/merged-plot.R cec2013/p1-stn-merged.RData 1.0
- Plot a dynamic merging MDSTN for multiple algorithms by
Rscript src/mdstn/merged-plot-gif.R <path to merged file .RData>
Example: Rscript src/mdstn/merged-plot-gif.R cec2013/p1-stn-merged.RData 1.0
We would like to express our gratitude to the authors of HillVallEA and CEC2013 Niching Benchmark which are the baseline which allows us to implement our new ideas; to the authors of STN for their graph modeling concepts.