Combining Inductive and Analytical Learning

Best Fits?

So, what does "best fits" mean?

User the old $error_D(d)$ definition—the proportion of examples from $D$ misclassified by $h$.
But, what about B?

6 of 32