Depends, itll probably help to get hired. If you are in a highly applied position, probably better to know different network architectures, and why they work in

Depends, itll probably help to get hired. If you are in a highly applied position, probably better to know different network architectures, and why they work in specific ways. For instance, Restnet versus VGG16, what was the innovation? Which optimizer might be better for the two? When to slow learning rate, when to speed it up? Theres an art to tuning and getting neural networks to work that goes beyond mathematical understanding. But some positions may require their employees to understand the archaic math behind machine learning, down PCA and its relationship to eigenvectors, methods of non differentiable optimization, etc