This data is for the paper "Search-based energy optimization of some ubiquitous algorithms" by Alexander Brownlee, Nathan Burles and Jerry Swan, to be published in IEEE Transactions on Emerging Topics in Computational Intelligence. The authors may be reached via sbr@cs.stir.ac.uk.

Please cite the above if this data is used in your own work.
This work is licensed under a Creative Commons Attribution 4.0 International License. See http://creativecommons.org/licenses/by/4.0/

======================================
Archive contents:

README.txt --- this document
Quicksort/training_arrays.tar --- the arrays used for the training phase of the quicksort experiments
Quicksort/makeArrays.py --- the arrays used for testing the optimised quicksorts were very large (100s of GB in total) so instead we have provided the Python script that was used to generate them. This uses a seeded RNG and should precisely replicate the data we used.
MLP/ --- the full set of evaluations from each of the MLP experiments (Section 5); grouped by data set (the data sets used in the experiments ship with WEKA and are available in arff format from the UCI machine learning repository)
    - eval@ files give the full structure with of each trained ANN
    - FUN files were output by jMetal and show the objective values (error and avg energy over 5 repeats) for each solution
    - VAR files were output by jMetal and show the variable values (AD params 0-3, then hidden layer size, backprop learning rate, backprop momentum) for each solution; each line corresponds to the same line in FUN
    - repeats shows the variable values with the corresponding 5 separate measured energy consumptions
Guava/TestCode --- the test classes used for the OOGI experiments (Section 6), these call the unit tests that ship with Guava v18
Guava/Outputs --- the optimised variants of the classes resulting from the OOGI experiments