The future of simple rules
The Future of Everything
00:00
The Importance of Regularization in Machine Learning
You don't have to go to massive data if you actually have some sort of understanding of the issue. In many situations, there may be 20 or 30 or 50 or 100 variables that are predictive but they're often only about five or six that really matter. And so what you do is essentially use what's known as regularization which just says, I'm only going to take the big effects. So in that case, you had a, your case studies, you generalized to a few potential rules and then you kind of check them on the big data to see if they hold up over a larger scale.
Transcript
Play full episode
