Training Clinically Actionable Machine Learning Models
Автор: Milan Toma
Загружено: 2025-11-14
Просмотров: 70
Описание:
You've trained your ML model and the accuracy is high. Time to celebrate? Not so fast. The most dangerous models aren't the ones that perform poorly; they're the ones that seem to perform brilliantly but crumble in production.
In this talk, I pull back the curtain on why this happens. This disconnect is almost always caused by methodological shortcuts (some intentional, some accidental) that create a false sense of security. I'll identify these critical errors, like data leakage and flawed validation, and provide a framework of disciplined practices to ensure the performance you calculate translates to genuine, real-world success.
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
