Snapshots from the Edge: Balancing Performance and Explainability in AI-enabled Medical Devices.

“We’ve always tried to use the simplest model possible that is fit for purpose. Your chance of at least partial explainability will increase when you adopt that mantra. And maybe there is a trade off where you have a model that has a slightly lower accuracy, but it has fewer input variables, it has a more transparent architecture.”

Snatching usable simplicity from the jaws of immense complexity … Jan Wolber, Digital Product Leader at GE Healthcare, on balancing the key trade-offs between performance and explainability in AI-enabled medical devices.