Question 1
Based on the above data, answer the given subquestions.
Deep Learning Quiz 2: 20 November 2022 (September 2022 term)
The IIT Madras BS Deep Learning (Deep Learning) Quiz 2 paper sat on 20 Nov 2022, in the September 2022 term: 12 questions for 40 marks in 120 minutes. Every question is below with its answer. Take it as a timed mock test to be marked, or read it through first.
- 12
- 40
- 120 min
- 9
- 2
- 1
Show answer
Correct answer: 4.4 (accepted within ±0.05)
Question 2
Based on the above data, answer the given subquestions.
Show answer
Correct answer: 0.89 (accepted within ±0.005)
Question 3
Based on the above data, answer the given subquestions.
Show answer
Correct answer: 0.905 (accepted within ±0.025)
Question 4
Based on the above data, answer the given subquestions.
Suppose we replace the sigmoid (logistic) activation function by ReLU and max pooling by min pooling (that is, output the minimum value from the input tensor). Update the parameters and enter the sum of updated parameter values?. The ReLU is defined as follows
Suppose we replace the sigmoid (logistic) activation function by ReLU and max pooling by min pooling (that is, output the minimum value from the input tensor). Update the parameters and enter the sum of updated parameter values?. The ReLU is defined as follows
(If your answer is 0.3345, enter it as 0.33)
Show answer
Correct answer: 0.1 (accepted within ±0.01)
Question 5
Based on the above data, answer the given subquestions.
The dimension of C1 = 101 × 101 × 100. Suppose we use 30 kernels (filters) of shape 3 × 3, with stride 1 and no zero padding, to generate the convolution layer 2. Then, enter the number of parameters (including bias) in the convolution layer-2 (If your answer is 1050,enter 1050)
Show answer
Correct answer: 27030 (accepted within ±1)
Question 6
Based on the above data, answer the given subquestions.
We know that the shape of the flattened layer is N ×1. What is the value of N?(If your answer is 1050,enter 1050)
Show answer
Correct answer: 17280 (accepted within ±0.5)
Question 7
Based on the above data, answer the given subquestions.
Suppose we apply three 1D kernels after the pooling layer, P2, and then flatten the final output to the shape of N × 1. What is the value of N now?
Show answer
Correct answer: 1728 (accepted within ±0.5)
Question 8
Based on the above data, answer the given subquestions.
Show answer
Correct answer: 0.0115 (accepted within ±0.0035)
Question 9
Based on the above data, answer the given subquestions.
Show answer
Correct answer: 2 (accepted within ±0.1)
Question 10
Based on the above data, answer the given subquestions.
The statement “Using sigmoid activation function for a neuron in a neural network always shrinks the gradient passing through it during backpropagation ” is
TRUE
FALSE
Show answer
Correct answer
TRUE
Question 11
The diagram below shows the contours of a loss surface generated by the following model,
The diagram below shows the contours of a loss surface generated by the following model,
where, m is the number of training samples, x1 and x2 are features and y is the corresponding ground truth. Assume that both features x1 and x2 are normalized to be in the same range.
The diagram below shows the contours of a loss surface generated by the following model,
where, m is the number of training samples, x1 and x2 are features and y is the corresponding ground truth. Assume that both features x1 and x2 are normalized to be in the same range.
Which of the following statement(s) is (are) true?
x1 is sparse, x2 is dense
x1 is dense, x2 is sparse
both x1 and x2 are sparse
both x1 and x2 are dense
Show answer
Correct answer
x1 is dense, x2 is sparse
Question 12
RMSProp
NADAM
NAG with cyclic learning rate
GD with exponential weight decay learning rate scheduler
Show answer
Correct answers
RMSProp
NADAM
NAG with cyclic learning rate
GD with exponential weight decay learning rate scheduler