GP Search configuration and Run

In order to run an experiment you have to execute (run) the project SearchAndReplaceGenerator.

SearchAndReplaceGenerator "Default run" configuration executes the project main by providing the configuration file configurations/examples_configuration.json

examples_configuration.json has been tuned for a quick demo execution.

You can adapt the file configurations/examples_configuration.json in order to execute the GP search with your preferred settings on your own datasets.

Interesting configuration details/parameters in the example:

• the selected strategy enforces elitism 0.4, elitism parameter is ignored
• "threads": "2" the number of threads for this GP search
• "generations": 30 the number of generations for each job
• "jobs": 4 the number of jobs

Furthermore, under "configurations" there are two separate configurations for the search and replace populations, i.e., each population have it's own "populationSize".

About the dataset, the "datasetContainer" part provides all the information on the csv dataset (that describes the search&replace task to perform). Under "datasetContainer", "path" is the path to the csv dataset, please note that the csv contains one example per line.

The "datasetContainer" section also provides the ranges (of dataset rows) to use in the training, validation and testing stage.

There are a lot of other parameters and details, but the parameter names should be self-explanatory.

At the end of each Run the program outputs the solution effectiveness on training, validation and testing sets, detailed informations are saved in the results.json file under the ./results sub-subfolder.