Deep Learning for Computer Vision, Quiz 2

May 2026 term, 16 Aug 2026, Set 1

Learning mode

Question 1

+1−0Single correct

Why do very deep plain CNNs (without skip connections) sometimes show higher training error than shallower CNNs?

A.
They always overfit because parameter count increases.
B.
Optimization becomes harder; skip connections make learning identity mappings easier and improve gradient flow.
C.
BatchNorm guarantees this never happens.
D.
Using sigmoid activations instead of ReLU always fixes it.

Deep Learning for Computer Vision, Quiz 2

May 2026 term, 16 Aug 2026, Set 1

Learning mode

+1−0Single correct

Why do very deep plain CNNs (without skip connections) sometimes show higher training error than shallower CNNs?

A.
They always overfit because parameter count increases.
B.
Optimization becomes harder; skip connections make learning identity mappings easier and improve gradient flow.
C.
BatchNorm guarantees this never happens.
D.
Using sigmoid activations instead of ReLU always fixes it.

Report a problem