Tag: Predictive Models

  • A contradiction in terms: AI interoperability in healthcare

    Most AI models can and will only ever be used at the institution(s) at which they were developed. The concept that an AI model is “shareable” — that a model developed at one place can be used at another — is generally a myth. This post explains why, and how the situation can be improved.

  • Avoiding False Alerts: Snoozing ≠ Laziness

    A simple approach can dramatically reduce false positives and annoying, redundant true negatives. Unfortunately, classic count-based metrics usually lead to the false conclusion that the approach wasn’t helpful! Our simple, novel approach solves that problem, enabling implementations with dramatically fewer false and useless alarms.

  • Shocking new discovery: Recall and Sensitivity are not the same!

    Classic statistics like sensitivity and specificity make assumptions that are usually false. That leads to serious problems. Our simple, novel approach provides the solution. Imagine this: My personal library contains 100 books, 50 with red bindings and 50 with blue bindings. I hide coins inside 20 of the books. 10 of the red books each…