This directory contains all you should need to prepare a sample submission for spatio-temporal challenges.

The code was tested with: Anaconda Python 3.7

Usage:

(1) If you are a challenge participant:

• The files sample_code_submission.zip contains a sample submission ready to go!

• The file README.ipynb contains step-by-step instructions on how to create a sample submission. At the prompt type: jupyter-notebook README.ipynb

Without running the jupyter-notebook, you can also directly:

• Modify sample_code_submission/model.py to provide a better model and win the challenge!

• Check your code in the same conditions it will be run on the platform, using the command:

python ingestion_program/ingestion.py sample_data sample_results ingestion_program sample_code_submission

• Download a larger dataset (called public_data) from the website of the challenge and re-test your code by replacing sample_data by public_data.

• To create a submission, zip the contents of sample_code_submission (without the directory, but with metadata)

(2) If you are a challenge organizer and use this starting kit as a template, ensure that:

• you modify README.ipynb to provide a good introduction to the problem and good data visualization

• sample_data is a small data subset carved out the challenge TRAINING data, for practice purposes only (do not compromise real validation or test data)

• the following programs run properly:

  python ingestion_program/ingestion.py sample_data sample_results ingestion_program sample_code_submission

  python scoring_program/score.py sample_data sample_results scoring_output

• the metric identified in metric.py is the metric used both to compute performances in README.ipynb and for the challenge.

• your code also runs within the Codalab docker (inside the docker, python 3.6 is called python3):

  docker run -it -v pwd:/home/aux codalab/codalab-legacy:py3

  DockerPrompt# cd /home/aux DockerPrompt# python3 ingestion_program/ingestion.py sample_data sample_result_submission ingestion_program sample_code_submission DockerPrompt# python3 scoring_program/score.py sample_data sample_result_submission scoring_output DockerPrompt# exit