Quiz Space

Deep Learning for Computer Vision · End Term · 10 May 2026 · January 2026 term

Deep Learning for Computer Vision End Term 10 May 2026 — Question 14

Question 14

+2 marksOne correct option

Vanishing gradients in vanilla RNNs occur mainly due to:

  1. A

    High deapth of the RNN layers

  2. B

    Repeated multiplication by Jacobians across timesteps

  3. C

    Too many convolution layers before RNN

  4. D

    Use of softmax

Show answer

Correct answer

  • B

    Repeated multiplication by Jacobians across timesteps

See the whole paperTake it as a mock test

Question 14 of 22 in the IIT Madras BS Deep Learning for Computer Vision (Deep Learning for Computer Vision) End Term paper sat on 10 May 2026, in the January 2026 term (Deep Learning For Computer Vision 06 May 26). It carries 2 marks.

More questions from this paper

  1. Q1A standard RGB image of size 256×256 is stored as 8-bit unsigned integers per channel. What is the total memory (in byt…
  2. Q2Which statement is correct?
  3. Q3You apply a 5×5 filter with stride 1 and no padding on a 32×32 image. Output size is:
  4. Q4The Sobel operator primarily estimates:
  5. Q5SIFT is designed to be robust mainly to:
  6. Q6In classical scale-space theory, increasing σ in Gaussian smoothing generally:
  7. Q7Backpropagation is best described as:
  8. Q8Which is a key effect of momentum?
  9. Q9L2 regularization on weights most directly encourages:
  10. Q10A key reason CNNs are parameter-efficient compared to fully connected nets for images is:
  11. Q11Stacking multiple 3×3 convolutions (stride 1) increases receptive field because:
  12. Q12In Faster R-CNN, the Region Proposal Network (RPN) outputs:
  13. Q13Which statement is correct ?
  14. Q15Select all correct statements:
  15. Q16Select all correct statements :
  16. Q17Select all correct statements:
  17. Q18Box A: top-left (0,0), bottom-right (8,8). Box B: top-left (4,4), bottom-right (12,12). Compute IoU as a decimal.
  18. Q19For a single query, suppose similarity with the positive is s+=2 and with two negatives are s1=1, s2=0. InfoNCE loss: L…
  19. Q20Input image: 64×64. Convolution: kernel 7×7, stride 2, padding 3. Compute output spatial size.
  20. Q21A fully connected network maps input 100 → hidden 50 → output 10. Ignore bias. What is the total number of weights?
  21. Q22Input feature map has M=64 channels. You apply depthwise 3×3 followed by pointwise 1×1 to produce N=128 channels. Ignor…