QA

Question 1

What does data normalization prevent?

Answer 1

reduce the risk of sigmoid saturation preventing minimization
stuck, premature convergence or undue increase of the estimated weights.

Question 2

T or F Data normalization prevents vanishing gradient

Answer 2

F

Question 3

For different task you always need different dataset T or F

Answer 3

T at least the label changes

Question 5

For small batch size the number of iterations required increases as the batch size increases t or f

Answer 5

F

Question 6

CNN rests on filter to reduce the image size t or f

Answer 6

F

Question 7

Activation maps are synonym of feature map t or f

Answer 7

T

Question 8

Hidden neurons are not active most of the time when the sparsity factor is approaching 0.5

Answer 8

T

Question 9

In Gan the discriminator maximizes the objective D(G(z)) =0 t or f

Answer 9

T

Question 10

In the training the gan learns the distribution of the images in the reference dataset t or f

Answer 10

T