Platforms
Codabench
ChaLearn uses a lot Codabench, an open source platform supporting code submission. The use of the public instance is free (you don’t need to install the platform on you own computers). Codabench allows you to design competitions with code submission (the code being evaluated in a way you define yourself, directly on the platform, with code execution). However, if you will use a lot of computational resources, you need to contribute “compute workers”, which are computers hooked up in the backend of Codalab and pick up jobs from a queue.
The development of Codabench and its ancestor Codalab, have been supported by ChaLearn. We are proud that Codalab has been cited two years in a row the #1 platform for scientific competitions (2023 report, 2024 report) by an independent entity MLContest. Codabench is quickly catching up (2025 report.)
Codabench quick start guide:
1) Download the example Iris competition bundle.
This is a ready-made sample competition bundle. Instructions to create your own are found on Github.
2) Go to the website:
https://competitions.codalab.org/
and create an account.
3) Click
> My Competitions
> Competitions I’m Running
> Create competition
> Upload a competition bundle
and navigate to find iris_bundle_19-05-13-16-15.zip that you downloaded in step 1.
Click “view” in the sentence: Congratulations! Your new competition is ready to view.
You are done, you should have your first competition. To know more, watch a brief TUTORIAL.
ChaLearn uses a lot Codabench, an open source platform supporting code submission. The use of the public instance is free (you don’t need to install the platform on you own computers). Codabench allows you to design competitions with code submission (the code being evaluated in a way you define yourself, directly on the platform, with code execution). However, if you will use a lot of computational resources, you need to contribute “compute workers”, which are computers hooked up in the backend of Codalab and pick up jobs from a queue.
The development of Codabench and its ancestor Codalab, have been supported by ChaLearn. We are proud that Codalab has been cited two years in a row the #1 platform for scientific competitions (2023 report, 2024 report) by an independent entity MLContest. Codabench is quickly catching up (2025 report.)
Codabench quick start guide:
1) Download the example Iris competition bundle.
This is a ready-made sample competition bundle. Instructions to create your own are found on Github.
2) Go to the website:
https://competitions.codalab.org/
and create an account.
3) Click
> My Competitions
> Competitions I’m Running
> Create competition
> Upload a competition bundle
and navigate to find iris_bundle_19-05-13-16-15.zip that you downloaded in step 1.
Click “view” in the sentence: Congratulations! Your new competition is ready to view.
You are done, you should have your first competition. To know more, watch a brief TUTORIAL.
Other platforms:
Kaggle: ChaLearn collaborators also organize challenges on Kaggle, which is the Machine Learning challenge platform of Google.
Tianchi: This is the Big Data platform of the Ailibaba Group, close in popularity to Kaggle, mostly used in China.
Causality Workbench: ChaLearn collaborators organized many challenges on this custom-made platform developed by MisterP. The challenges are still alive, but the platform is no longer developed and used for new challenges.
Many other platforms are available to organize Machine Learning challenges. See the Challenges page.
Kaggle: ChaLearn collaborators also organize challenges on Kaggle, which is the Machine Learning challenge platform of Google.
Tianchi: This is the Big Data platform of the Ailibaba Group, close in popularity to Kaggle, mostly used in China.
Causality Workbench: ChaLearn collaborators organized many challenges on this custom-made platform developed by MisterP. The challenges are still alive, but the platform is no longer developed and used for new challenges.
Many other platforms are available to organize Machine Learning challenges. See the Challenges page.
