Question 81 MSQ · 2.0 marks
4
3
2
0
1
A published solution is not available for this question yet.
Deep Learning · End Term · Sep 2023 · FN
Deep Learning · End Term · Sep 2023 FN
← Course papers · Start practice / exam
This page contains the reliably extracted subset, not the complete original paper.
Questions and published explanations below are available without starting a test. Some questions may not have a published solution yet.
0 / 17 published solutions
Question 82 MCQ · 3.0 marks
1
2
4
It oscillates and never converges
A published solution is not available for this question yet.
Question 83 NAT · 3.0 marks
A published solution is not available for this question yet.
Question 84 NAT · 3.0 marks
A published solution is not available for this question yet.
Question 85 MSQ · 5.0 marks
The loss surface is convex (with a global minimum)
The loss value oscillates over iterations
The loss value decreases consistently over iterations
The loss value increases consistently
A published solution is not available for this question yet.
Question 86 MSQ · 5.0 marks
Definitely, the loss value at the end of the training is zero
The neuron in the hidden layer might be experiencing a vanishing gradient
problem
The loss could have converged to the local minimum after a finite number of
iterations
The loss value at the end of the training may not necessarily be zero
A published solution is not available for this question yet.
Question 87 MSQ · 3.0 marks
Suppose that a neural network has millions of parameters (weights and biases). A team decides to
use an optimization algorithm with a learning rate scheme that is **local** to each parameter in the
network. Moreover, the learning rate changes in each iteration such that it should decrease on the
steep surface and increase on the gentle surface. Which of the following optimization algorithms
satisfy the team’s requirements?
GD with an exponentially decaying learning rate scheduler
AdaGrad
AdaM
NADAM
RMSProp
SGD with line search
A published solution is not available for this question yet.
Question 88 MSQ · 3.0 marks
Suppose we divide the available training samples into mini batches of size 32 to train a model with
mini-batch gradient descent. Assume that we have 33 different machines to train the model. One
out of 33 machines acts as a master machine. The actual weight update happens in the master
machine. The master machine can send one sample for the rest of the machines along with a copy
of the model in its current state to compute the gradients. We call this entire set-up parallelization.
Which of the following deep learning architectures can be trained in parallel then?
Fully connected Feed forword neural network
Convolutional Neural network
Reccurent Neural Network
Transformers
A published solution is not available for this question yet.
Question 89 MSQ · 4.0 marks
A published solution is not available for this question yet.
Question 90 NAT · 5.0 marks
A published solution is not available for this question yet.
Question 91 NAT · 2.0 marks
Based on the above data, answer the given subquestions.
How many number of parameters are there in the network (excluding bias)?
A published solution is not available for this question yet.
Question 92 MCQ · 1.0 marks
Based on the above data, answer the given subquestions.
The word representation for the 2-nd word in the vocabulary corresponds to
A published solution is not available for this question yet.
Question 93 NAT · 2.0 marks
Based on the above data, answer the given subquestions.
What is the size of the vocabulary |V |?
A published solution is not available for this question yet.
Question 94 MCQ · 2.0 marks
Based on the above data, answer the given subquestions.
Suppose we consider the three words (wood,woodchuck, much). Assume we use one-hot encoded
vector representation for all these words. The statement that,“The cosine similarity between the
pair (wood,woodchuck) is greater than the pair (wood,much)” is
True
False
A published solution is not available for this question yet.
Question 95 NAT · 2.0 marks
Based on the above data, answer the given subquestions.
How many parameters (including bias) are there in the network?
A published solution is not available for this question yet.
Question 96 NAT · 2.0 marks
Based on the above data, answer the given subquestions.
For the given input sentence containing 15 words, how many sequential time steps does RNN take
to make a final prediction?
A published solution is not available for this question yet.
Question 97 NAT · 3.0 marks
Based on the above data, answer the given subquestions.
Suppose we train the model using the BPTT algorithm for 100 iterations. In each iteration, we feed
the input sentence and make a prediction, compute the loss, back-propagate through time and
update the parameter. How many times does the parameter matrix W get updated over 100
iterations?
A published solution is not available for this question yet.