4 ms·
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.
by TuringNYC 9y ago
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.
- madenine 9y agoI imagine getting samples/training sets that don't have some hefty source of bias is really tough in medicine, and probably leads to over-fitting.