ParentFull threadTuringNYC·I'm honestly amazed at how some papers (esp in medicine) magically train deep nets using 1000 or 2000 cases and get absurdly high 90s AUC.View on HN