Deep Learning
Residual Connection
y = x + F(x), giving ∂y/∂x = I + ∂F/∂x - a gradient path that passes through no weight matrices.
Why interviewers ask about it
ResNet's insight was that deep plain networks had higher *training* error: an optimisation failure, not overfitting. Every transformer block uses the same idea.
Related terms
This term is part of the free AI/ML Engineer interview preparation module - browse the full glossary for every definition.