Deep Learning Concepts Assessment

Page 1 sur 16Lecteur de document UniversityLib

Deep Learning Concepts Assessment

Deep Learning and Neural Networks · exam

Voir tous les documents en intelligence artificielle et données

Question 1

Correct

Note de 1,00 sur 1,00

To which of these tasks would you apply a many-to-one RNN architecture? (Check all that apply).

A. Machine Translation (input a piece of text and the output is a translated text)

B. Sentiment classication (input a piece of text and output a 0/1 to denote positive or negative

sentiment)

C. Gender recognition from speech

D. Image classication

Votre réponse est correcte.

Les réponses correctes sont :

Sentiment classication (input a piece of text and output a 0/1 to denote positive or negative sentiment),

Gender recognition from speech

Question 2

Correct

Note de 1,00 sur 1,00

Consider this RNN: This specic type of architecture is appropriate when:

A. T = Ty

x

B. T = T =1

x

y

C.

T < T

y

x

D. T > T

y

x

Votre réponse est correcte.

La réponse correcte est :

T = Ty

x

Question 3

Correct

Note de 1,00 sur 1,00

You have nished training a language model RNN and are using it to sample random sentences, as

follows:

What are you doing at each time step t:

A.

(1) The highest probability word of the time step t as

is picked from output probabilities,

(2) the ground truth word from the training set is passed to the next step t+1.

B. (1) Randomly sample a chosen word according to the given output probabilties  of the RNN  for the

time step

, (2) the selected word is passed to the next step t+1.

C.

(1) Use the probabilities given by the RNN to sample a word for the time step

, (2) the

ground truth word from the training set is passed to the next step t+1.

D. (1) Pick the highest probability from the output probabilities of the RNN  for the time step

, (2) the selected word is passed to the next step t+1.

Votre réponse est correcte.

La réponse correcte est :

(1) Randomly sample a chosen word according to the given output probabilties  of the RNN  for the time step

, (2) the selected word is passed to the next step t+1.

Question 4

Correct

Note de 1,00 sur 1,00

Which of these is the most likely cause of problems for RNNs:

A. Vanishing gradients.

Publicité

B. Exploding gradients.

C. Activations are too large.

D. Dicult to process longer sequences.

Votre réponse est correcte.

Les réponses correctes sont :

Exploding gradients.,

Vanishing gradients.,

Dicult to process longer sequences.

Question 5

Correct

Note de 1,00 sur 1,00

The Gated Recurrent Units (GRU) helps avoiding vanishing/exploding gradients

Where:

Check all that apply.

A. The nal output state h is a combination of h  and ϕ (W x +U (r ⊙h )+b ) via z and it is the

t−1

h t

h t

t−1

t

t

h

h

representation of the hidden state h

t -1

B. The nal output state h is a combination of h  and ϕ (W x +U (r ⊙h )+b ) via z and it is the

t−1

h t

h t

t−1

t

t

h

h

representation of the input x

t

Votre réponse est correcte.

Les réponses correctes sont :

The nal output state h is a combination of h  and ϕ (W x +U (r ⊙h )+b ) via z and it is the

t−1

h t

h t

t−1

t

t

h

h

representation of the input x ,

t

The reset gate r  is used to decide how much of the past information to forget.

t

Question 6

Correct

Note de 1,00 sur 1,00

We talked about “parameter sharing” as a benet of using convolutional networks. Which of the

following statements about parameter sharing in ConvNets are true? (Check all that apply.)

A.

It allows gradient descent to set many of the parameters to zero

B. It reduces the total number of parameters

C. It allows parameters learned for one task to be shared even for a dierent task.

D. It allows a feature detector to be used in multiple locations throughout the whole input

Publicité

image/input volume.

Votre réponse est correcte.

Les réponses correctes sont :

It reduces the total number of parameters,

It allows a feature detector to be used in multiple locations throughout the whole input image/input volume.

Question 7

Correct

Note de 1,00 sur 1,00

Which of the following do you typically see in a ConvNet? (Check all that apply.)

A. Multiple POOLING layers followed by a CONV layer.

B. Multiple CONV layers followed by a POOLING layer.

C. Fully-Connected layers in the rst few layers

D. Fully-Connected layers in the last few layers.

Votre réponse est correcte.

Les réponses correctes sont :

Multiple CONV layers followed by a POOLING layer.,

Fully-Connected layers in the last few layers.

Question 8

Correct

Note de 1,00 sur 1,00

In order to be able to build very deep networks, we usually only use pooling layers to downsize the

height/width of the activation volumes while convolutions are used with “valid” padding. Otherwise, we

would downsize the input of the model too quickly.

A. True

B. False

Votre réponse est correcte.

La réponse correcte est :

False

Question 9

Correct

Note de 1,00 sur 1,00

Suppose you have an input volume of dimension 64x64x32. How many parameters would a single 1x1

convolutional lter have (including the bias)?

A. 32

B. 33

C. 4096

D. 4097

Votre réponse est correcte.

La réponse correcte est :

33

Question 10

Correct

Note de 1,00 sur 1,00

You have an input volume that is 15x15x8, and pad it using “pad=2.” What is the dimension of the

resulting volume (after padding)?

A. 17x17x16

B. 19x19x8

C. 19x19x16

D. 17x17x8

Votre réponse est correcte.

La réponse correcte est :

19x19x8

Question 11

Partiellement correct

Note de 0,50 sur 1,00

Which of these statements about mini-batch gradient descent do you agree with?

A. You should implement mini-batch gradient descent with an explicit for-loop over dierent mini-

batches, so that the algorithm processes mini-batches at each iteration.

B. Training one epoch (one pass through the training set) using mini-batch gradient descent is faster

than training one epoch using batch

C. One iteration of mini-batch gradient descent (computing on a single mini-batch) is faster than

Publicité

one iteration of batch gradient descent.

Votre réponse est partiellement correcte.

Vous en avez sélectionné correctement 1.

Les réponses correctes sont :

You should implement mini-batch gradient descent with an explicit for-loop over dierent mini-batches, so

that the algorithm processes mini-batches at each iteration., One iteration of mini-batch gradient descent

(computing on a single mini-batch) is faster than one iteration of batch gradient descent.

Question 12

Partiellement correct

Note de 0,33 sur 1,00

Which gure correspond to the right Padding

Votre réponse est partiellement correcte.

Vous en avez sélectionné correctement 1.

La réponse correcte est :

→ Valid,

SameTotalValid → Total,

→ Same

Question 13

Correct

Note de 1,00 sur 1,00

In a neural network, weights are updated as follows:

Take a batch of training data and perform forward propagation to compute the loss.

Backpropagate the loss to get the gradient of the loss with respect to each weight.

Use the gradients to update the weights of the network.

Votre réponse est correcte.

La réponse correcte est :

Take a batch of training data and perform forward propagation to compute the loss. → 1,

Backpropagate the loss to get the gradient of the loss with respect to each weight. → 2,

Use the gradients to update the weights of the network. → 3

123Question 14

Incorrect

Note de 0,00 sur 1,00

Consider the two following random arrays "a" and "b":


   a = np.random.randn(2, 3) # a.shape = (2, 3)

    b = np.random.randn(2, 1) # b.shape = (2, 1)

    c = a + b

What will be the shape of "c"? (e.g., (x,y) )

Réponse :

(2,3)

La réponse correcte est : (2, 3)

Question 15

Incorrect

Note de 0,00 sur 1,00

Recall that np.dot(a,b) performs a matrix multiplication on a and b, whereas a*b performs an element-

wise multiplication.

Consider the two following random arrays "a" and "b":


    a = np.random.randn(12288, 150) # a.shape = (12288, 150)

    b = np.random.randn(150, 45) # b.shape = (150, 45)

    c = np.dot(a, b)

What is the shape of c?

Réponse :

(2,3)

La réponse correcte est : (12288, 45)

Question 16

Correct

Note de 1,00 sur 1,00

Which of the following statements is true?

A. The deeper layers of a neural network are typically computing more complex features of the

input than the earlier layers.

B. The earlier layers of a neural network are typically computing more complex features of the input

than the deeper layers.

Votre réponse est correcte.

La réponse correcte est :

The deeper layers of a neural network are typically computing more complex features of the input than the

earlier layers.

Publicité

Question 17

Correct

Note de 1,00 sur 1,00

During forward propagation, in the forward function for a layer l you need to know what is the activation

function in a layer (Sigmoid, tanh, ReLU, etc.). During backpropagation, the corresponding backward

function also needs to know what is the activation function for layer l, since the gradient depends on it.

True/False?

A. False

B. True

Votre réponse est correcte.

La réponse correcte est :

True

Question 18

Correct

Note de 1,00 sur 1,00

Which of the following are true? (Check all that apply.)

A. Decreasing the size of a neural network generally does not hurt an algorithm’s performance, and it

may help signicantly.

B. Increasing the size of a neural network generally does not hurt an algorithm’s performance, and

it may help signicantly.

C. Decreasing the training set size generally does not hurt an algorithm’s performance, and it may help

signicantly.

D. Increasing the training set size generally does not hurt an algorithm’s performance, and it may

help signicantly.

Votre réponse est correcte.

Les réponses correctes sont :

Increasing the training set size generally does not hurt an algorithm’s performance, and it may help

signicantly.,

Increasing the size of a neural network generally does not hurt an algorithm’s performance, and it may help

signicantly.

Question 19

Partiellement correct

Note de 0,33 sur 1,00

Match the given situation with the appropriate category

Agent is only presented with the inputs themselves, and aims to nd structure in

these inputs

Agent is not presented with target outputs, but is given a reward signal, which it

aims to maximize

Agent is presented with examples of inputs and their target outputs

Votre réponse est partiellement correcte.

Vous en avez sélectionné correctement 1.

La réponse correcte est :

Agent is only presented with the inputs themselves, and aims to nd structure in these inputs →

Unsupervised Learning,

Agent is not presented with target outputs, but is given a reward signal, which it aims to maximize →

Reinforcement Learning,

Agent is presented with examples of inputs and their target outputs → Supervised Learning

Supervised LearningReinforcement LearningUnsupervised LearningQuestion 20

Incorrect

Note de 0,00 sur 1,00

Match the given convolutional lter with the right appropriate layer (three layers):

Votre réponse est incorrecte.

La réponse correcte est :

→ First Layer,

→ Third Layer,

Third LayerSecond LayerFirst Layer → Second Layer

◄ Test Evaluation #2

Aller à…

Test Evaluation #4 ►