/fun/forest
Model Escalation
The path from "let's try a simple model" to "just use a linear regression".
1. Decision treeSimple. Interpretable. Underfitting.
2. Random forestOh, decision trees aren't good enough for you?
3. XGBoostGradient boosted. State of the art. 47 hyperparameters.
4. Neural networkOverkill. But it works. Sometimes.
5. Linear regressionYou knew it all along. Occam was right.