• LeFrog@discuss.tchncs.de
    link
    fedilink
    arrow-up
    4
    ·
    edit-2
    2 months ago

    Thx for your answer, makes sense! I was already wondering how the recommendations are generated. Maybe you can put this in the readme as well? I think that will make it more clear what a user can expect.

    • Mubelotix@jlai.luOP
      link
      fedilink
      arrow-up
      3
      ·
      2 months ago

      Great idea, thank you for your feedback! I generated embeddings for repositories using a SVC model trained on a dataset containing almost all stars of Github. Recommendations are produced by finding the nearest neighbors in the resulting vectorial space. I might be improving this even further by including forks and watcher data in the future, but it’s becoming expensive to train