Introduction to Deep Learning

Introduction to Deep Learning 

1 of 2outlined_flag

Which of the following pair of DNNs are similar? 

High cost hardware

Availability of computing power check

Unavailability of good amount of data

Option Availability of computing power is Correct

Q2 of 2outlined_flag

According to the Hype cycle, Deep Learning is in which phase?

Innovation trigger

Plateau of Productivity

Trough of Disillusionment

Peak of Inflated Expectations check

Q1 of 2outlined_flag

Which of the following are responsible for the nonlinear output of an artificial neural network?

Activation functions check

Weights

Biases

Option Activation functions is correct

Q2 of 2outlined_flag

During the learning phase of an artificial neural network, the weights and biases are __________.

altered during backward propagation check

altered during forward propagation

never altered

Option altered during backward propagation is correct

I want to classify an RGB image, size 32x32 using a Simple Neural Network (Not a CNN). What will be the number of neurons in input layer?

32 X 32check

32 X 32 X 2

32 X 32 X 3 

None of these

Option 32 X 32 X 3 is Correct

Q2 of 2outlined_flag

If we apply a max pool with 2 X 2 filter and stride of 2 on a 4 X 4 matrix, what will be the dimension of the resultant matrix?

4 X 4

2 X 2 X 2

3 X 3

2 X 2 check

Option 2 X 2 is Correct

Q1 of 2outlined_flag

Long Short Term Memory (LSTM) Network is a type of?

Static Model

Temporal Model check

Autoencoders

Option Temporal Model is correct

Explanation :

Correct!

Q2 of 2outlined_flag

Which of the following is a better model to choose for predicting sock market prices? 

CNN

RNN check

Autoencoders

Option RNN is correct

Explanation :

New data is passed as an input in RNN

Q1 of 1outlined_flag

Which of the following is TRUE regarding Autoencoder?

Used for removing patches from the images

It is an unsupervised way to learn about the input

Encoding and Decoding phases are the mirror layers

All of the above

Submit

Q1 of 1outlined_flag

Which of the statements is INCORRECT regarding Deep Learning?

Can be used for loan underwriting

Cannot be used for conservation of energy check

Can be used for predicting the best stock to invest in

Explanation :

DL can be used for efficient utilization of resource

Q1 of 2outlined_flag

Which of the following is NOT developed by Google?

 Tensorflow

Pytorch check

Keras

None of these

Explanation :

Pytorch is developed by Facebook

Q2 of 2outlined_flag

Which of the following is JAVA based?

Theano

Tensorflow

DL4J check

pytorch

Explanation :

Can only be used with Java

Self Assessment - Deep Learning

Section Score: 10/10

Correct

Incorrect

Unattempted

Q1 of 10

Usually RNN is not used for which of the following tasks?

Stock price detection

Dimensionality reduction

Speech recognition

Q2 of 10

Which of the following pair of DNNs are similar? 

CNN - RNN

RNN - LSTM

LSTM - CNN

Autoencoder - LSTM

Q3 of 10

Which of the following frameworks is primarily developed by Facebook AI research group?

DL4J

Tensorflow

Pytorch

Keras

Q4 of 10

Arrange the steps in the deep learning process in the correct order --

1. Splitting the data in training and testing datasets

2. Model validation

3. Gathering data

4. Applying traditional machine learning algorithm

5. Applying deep learning algorithm

i, iii, ii, v, iv

ii, iii, iv, v

iii, i, v, ii

i, iii, v, iv

Q5 of 10

Assume we have stock market data X1, X2, X3, X4 for day1, day2, day3, day4 respectively. In RNN, the past day's data is used as:

a weight

a bias

an input

All of the above

Q6 of 10

Which of the following tasks a Deep Learning model cannot do yet?

Identify a fraudulent transaction

Translate a text from one language to another

Innovate a novel business idea

Detect a disease from medical data

Q7 of 10

When should we NOT use Autoencoders? 

To predict tomorrow's weather

To reduce the dimension of the data

To remove noise from the data

Q8 of 10

Which of the following is the correct order of working in CNN?

1. Flattening

2. Convolution

3. Pooling

i,ii,iii

iii,ii,i

ii,i,iii

ii,iii,i

Q9 of 10

If we apply max pool with 2 X 2 filter and stride as 2 on 8 X 8 matrix, the dimension of resultant matrix will be?

2 X 2

4 X 4

6 X 6

8 X 8

Q10 of 10

When should we use Deep Learning?

When we have high end infrastructure

When we have lack of domain knowledge for feature introspection

All of the above