Benchmark comparison experiments for the paper "A review of algorithmic approaches for cell culture media optimization"
- Create conda environment
conda env create -f environment.yml
- Follow instructions to install COCO (for test functions) provided here: https://github.com/numbbo/coco
- Edit the problem constants in
run_experiment.pyto run the experiment of choice
- dim : dimension of problem, choices are 5, 20, or 40
- population : population size of each generation. For dim=5, this will be ignored if CCD or BBD DOE is chosen, and replaced by the default number as determined by the respective DOE methods
- iteration : number of iterations
- offset : offset value for function input, i.e. f(x) -> g(x - offset), where g is the original BBOB test function. Purpose of offset is prevent the minima solutions at x = 0, which is present in standard DOEs
- noisy : if to add noise to function
- noise_ratio : controls variance for gaussian noise, range [0 - 1]. y = y + N(0, y*noise)
- replicates : number of replicates to perform per experiment
- test_set : an interable to specify test function by ID (1-24). E.g. If test all: range(1,25), if only 1: [1]
- methods : list of all methods used in this experiment. If a subset of methods is to be used, override this list with a custom list
- In the
optimizer_expconda environment, run script:
python run_experiment.py
- Find results stored in corresponding results folder