Slow algorithms: More models

- [Narrator] The third reason why…some algorithmic approaches are going to be slower…is we literally build more models in some cases.…So for instance, going all the way back to the 80's…it was proposed that we could build…more accurate decision trees if we simply built a bunch…of them and then averaged.…That's called bagging, but over the years a number…of new algorithms have been developed…that capitalize in this same idea.…Somewhat famously there's one called random forests.…

More recently, there's a fancier one called XGBoost,…but all of these do that same basic notion.…They build lots of trees and then average them.…So think about this now.…If you're going to build 50 trees and average them,…you know that your computational time…is going to be about 50 times more.…That can start to catch up with you.…There's a completely different sense…in which we could talk about building more models.…Let me give you an example.…Years ago I was running an analytics team…and one of the big projects that we did…was building models for each SKU so each bike model…