Aug 9, 2017 · 1 min read
Nice review of model selection.
With k predictors, M1 creates k models, M2 creates k-1 models, etc. Summing M1 through Mk gives k+(k-1) + … 1 = (k+1)k/2. It’s O(k²), not k².
In Fernando’s case, k = 5, so he will create 15 different models. Even less if he has a stopping criterion.
