Repository for generating policies in the OLL-OneMax environment

What this means is, we are creating dynamic parameter configurations for the (1 + (lambda, lambda)) genetic algorithm with the aim of making it search OneMax landscapes faster.

To build the conda environment:

./.deploy/BUILD

To start generating

conda activate .deploy/conda_environment
python ./.deploy/range.py
./.deploy/RUN_GENERATE config/continuous/affront.yaml

The output will be under ./compute/<your_hostname>/continuous/. The output is a collection of .db files, viewable with DB Browser for SQLite.

You can visualize the .db files using this repository.

One .db file corresponds to one setting, according to which the configurator is run. The configurator will output a configuration of the (1 + (lambda, lambda)) algorithm. But the configurator itself must be adjusted. This is accomplished via .yaml files in the ./config/ subtree.

One such adjustment and run of the configurator is called an experiment. Therefore, one .db file is an experiment, in which the configurator is creating more and more policies for the (1 + (lambda, lambda)) algorithm, as it trains searching OneMax landscapes. Currently, in one experiment, we only train on one OneMax landscape, given by a fixed dimensionality.