Deep Learning Practice · End Term · 13 Sept 2026 · May 2026 term · Set 1

Question 25: Compared with the sigmoid activation, which of the follo…

Question 25

+2 marksOne or more correct options

Compared with the sigmoid activation, which of the following are correct reasons for preferring the ReLU activation in deep CNNs?

Select all that apply.

A
For positive inputs, ReLU does not saturate, so it avoids the gradient-killing that occurs in the flat tails of the sigmoid.
B
ReLU is preferred because it bounds every activation to the interval [0, 1], preventing activations from growing large.
C
ReLU is cheaper to evaluate — a simple thresholding at zero — whereas sigmoid/tanh require expensive exponentials.
D
The sigmoid's derivative peaks at only 0.25, so multiplying many such small factors across layers pushes gradients toward zero (vanishing gradient).

Show answer

Correct answers

A
For positive inputs, ReLU does not saturate, so it avoids the gradient-killing that occurs in the flat tails of the sigmoid.
C
ReLU is cheaper to evaluate — a simple thresholding at zero — whereas sigmoid/tanh require expensive exponentials.
D
The sigmoid's derivative peaks at only 0.25, so multiplying many such small factors across layers pushes gradients toward zero (vanishing gradient).

See the whole paper Take it as a mock test

Question 25 of 31 in the IIT Madras BS Deep Learning Practice (Deep Learning Practice) End Term paper sat on 13 Sept 2026, in the May 2026 term (Deep Learning Practice 13 Sep 26 (Session 2)). It carries 2 marks.

Question 25: Compared with the sigmoid activation, which of the follo…

Question 25

More questions from this paper