Professional Documents
Culture Documents
Dear All,
Read all the instructions carefully.
This quiz has been timed for 20min.
DO NOT EDIT the FORM_TIMER_UNIQUE_IDENTIFIER field. It will get updated automatically.
Take screenshots of your responses in case you face issues during submission.
Thank you.
* Required
1. Email address *
2. PRN *
Questions
3. The following 3 questions are based on this image. Given a vector as follows, what is
its L1 norm? (NOTE: Simply mention the computed number and for floating point
numbers, round off the number upto 3 decimal places) *
6. Given 60000 grayscale images of size 28 x 28, what would be the shape of tensor
required? *
(60000, 28*28)
Questions (contd)
vector of partial derivatives of the loss score (or empirical loss) with respect to every
weight in the network
vector of derivatives of the loss score (or empirical loss) with respect to every weight in
the network
vector of the loss score (or empirical loss) computed with respect to every weight in the
network
Questions (contd)
11. Consider the steps involved in the gradient descent process of optimization, which of
the following options do you think should be an ideal size for a batch? *
13. Now considering the same graph write the output of grad(loss_val, w) [NOTE:
Follow the notations as mentioned in the previous question] *
14. Thus the chain rule in this backward graph says that you can obtain the derivative
of a node with respect to another node by *
multiplying the derivatives for each edge along the path linking the two nodes
adding the derivatives for each edge along the path linking the two nodes
multiplying the derivatives for every intermediate node along the path linking the two
nodes
15. Read the following question carefully *
-2.808
-0.648
9.36
-3.6
-1.808
16. Continuing with the example above, *
-1.249
-2.28
3.369
-2.378
-0.249
Forms