If you identified that your model has high bias do you assume that it is underfit and start adding features or decrease lambda?

Likewise if you identify that your model has high variance do you assume that your model is overfitting and you reduce features or increase Lambda?


Initially, set lambda to 0. Don’t increase lambda until you already have overfitting.

Reducing features is rarely a good idea, because you’re throwing away data that might be important. Increase lambda instead.

Got it thank you!