Model Selection & Generalization

The modeling phase is the heart of the interview, and it is graded on judgment, not name-dropping. Interviewers want to see you start simple, compare real alternatives on real axes, commit to one, and then defend its details under probing. This page covers that sequence, the architectures worth knowing, loss function design, and the generalization toolkit (overfitting, capacity, transfer learning, regularization) that backs it all up.