home.social

#maxent — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #maxent, aggregated by home.social.

  1. Students often have a hard time conceptualizing maximum entropy (#MaxEnt) methods. Here's a little two-application demo (species distribution modeling and NLP) [written by #ClaudeAI] that helps make the construction of these distributions a little clearer.
    tpavlic.github.io/asu-bioinspi

  2. I'm at #LifeCLEF2023 (online). Interesting to hear about #GeoLifeCLEF, a difficult challenge: predicting which plant species are present, from satellite images... where the training data are "positive-only": observations are listed, but no "un-observations"! Apparently #maxent is still a good strong baseline, though deep learning can do a bit better.

  3. I'm at #LifeCLEF2023 (online). Interesting to hear about #GeoLifeCLEF, a difficult challenge: predicting which plant species are present, from satellite images... where the training data are "positive-only": observations are listed, but no "un-observations"! Apparently #maxent is still a good strong baseline, though deep learning can do a bit better.

  4. I'm at #LifeCLEF2023 (online). Interesting to hear about #GeoLifeCLEF, a difficult challenge: predicting which plant species are present, from satellite images... where the training data are "positive-only": observations are listed, but no "un-observations"! Apparently #maxent is still a good strong baseline, though deep learning can do a bit better.

  5. I'm at #LifeCLEF2023 (online). Interesting to hear about #GeoLifeCLEF, a difficult challenge: predicting which plant species are present, from satellite images... where the training data are "positive-only": observations are listed, but no "un-observations"! Apparently #maxent is still a good strong baseline, though deep learning can do a bit better.

  6. I'm at #LifeCLEF2023 (online). Interesting to hear about #GeoLifeCLEF, a difficult challenge: predicting which plant species are present, from satellite images... where the training data are "positive-only": observations are listed, but no "un-observations"! Apparently #maxent is still a good strong baseline, though deep learning can do a bit better.