The Open Algorithm Selection Challenge (OASC) will take place in Summer 2017. To this end, we provide here some simple scripts which allow newcomers to read the data into pandas and show an example how to generate an output file in the required format.
The example requires Python 3.4 or later. To install the required dependencies please call:
pip install -r requirements.txt
(Most requirements are already satisfied by using Anaconda.)
The example script simpy computes the single best algorithm across all training instances and predicts it for each test instance.
To call the script, please call:
python oasc_starterkit/single_best.py --train_as example_files/SAT11-INDU-TRAIN/ --test_as example_files/SAT11-INDU-TEST/
Please have a look in the extensively documented code for some explanations.
In the end, the scripts writes the file results.json
to disk
which could be submitted in this format to the competition.
In validation/
, we provide a script to validate your results files
on known test data. Please note that all files of the test scenario has to be provided for this script.
Example call:
python validation/validate_cli.py --result_fn results.json --test_as example_files/SAT11-INDU-TEST/ --train_as example_files/SAT11-INDU-TRAIN/
Add "." to your PYTHONPATH to avoid import errors, e.g.,
export PYTHONPATH=.//:$PYTHONPATH
Marius Lindauer lindauer@cs.uni-freiburg.de