Paper Implementations grouped by framework
Percentage of published papers that have at least one code implementation
Frameworks: Repositories are classified by framework by inspecting the contents of every GitHub repository and checking for imports in the code. We limit to repositories that are implementations of papers. The date axis is the date the repository was created (NOTE: pytorch/tf support might have been added later - which explains why some repositories originally started in 2014/2015 are marked as pytorch/tf).
Code Availability: For every open access machine learning paper, we check whether a code implementation is available on GitHub. The date axis is the publication date of the paper. We include both official and community implementations.