Another Machine Vision/Object Rep Learning Paradigm

› Forums › Personal Topics › Unbidden Thoughts › Another Machine Vision/Object Rep Learning Paradigm

This topic contains 1 reply, has 1 voice, and was last updated by  josh December 10, 2020 at 10:14 pm.

  • Author
    Posts
  • December 10, 2020 at 10:14 pm #75651

    josh

    Another variation for testing probability judgments of a ML – e.g. for a question “Is this a dog1, dog2, new dog, or something else? What probabilities do you assign?” – is to train the ML to optimize performance on a betting task with a Scoring Rule or setting where a Kelly Criterion applies. That adds an additional layer of effort, but yields an objective way of talking about the ML’s resultant likelihood function.

  • Author
    Posts

You must be logged in to reply to this topic.