Using AI and ML to Help Humans, Not Replace Them - ML 071
Adventures in Machine Learning
00:00
Machine Learning - Is Cross-Validation Worthwhile?
If you're just saying, hey, the accuracy was 89%, we're good to go and you don't monitor it. You might be surprised three, six, nine, 12 months later when you've started to see the sort of drift in multiple points. The data you train on is a sample from a certain time slice in production. If you're not aware of how the distribution of your data changes over time, you can run into some awkward embarrassing issues very quickly.
Transcript
Play full episode
