DPL-Course1-_Course4 Exam
Which of the following are some aspects in which AI has A-Web searching and advertisement.
transformed business?
AI has helped to make a fit between services or results and consumers or queries.
A-Web searching and advertisement.
B-Creating an AI-powered society.
C-Eliminating the need for health care services.
D-AI has not been able to transform businesses.
Which of the following play a major role to achieve a very A-Large amounts of data.
high level of performance with Deep Learning algorithms?
A-Large amounts of data. C-Large models
B-Better designed features to use. D-Deep learning has resulted in significant improvements in important applications
C-Large models. such as online advertising, speech recognition, and image recognition.
D-Deep learning has resulted in significant improvements
in important applications such as online advertising,
speech recognition, and image recognition.
E-Smaller models.
Recall the diagram of iterating over different ML ideas. A-Experiments finish faster, producing better ideas through increased iteration
Which of the stages shown in the diagram was improved tempo.
with the use of a better GPU/CPU?
C-Some algorithms are specifically designed to run experiments faster.
A-Experiments finish faster, producing better ideas through
increased iteration tempo.
B-Without better hardware, there is no way to train models
faster.
C-Some algorithms are specifically designed to run
experiments faster.
D-With larger datasets, the iteration process is faster.
When experienced deep learning engineers work on a new False
problem, they can usually use insight from previous
problems to train a good model on the first try, without Finding the characteristics of a model is key to having good performance. Although
needing to iterate multiple times through different models. experience can help, it requires multiple iterations to build a good model.
True/False?
Images for cat recognition is an example of "structured" False
data, because it is represented as a structured array in a
computer. Images for cat recognition are examples of "unstructured" data.
True/False?
A demographic dataset with statistics on different cities' False
population, GDP per capita, and economic growth is an
example of "unstructured" data because it contains data A demographic dataset with statistics on different cities' population, GDP per capita,
coming from different sources. and economic growth is an example of "structured" data in contrast to image, audio
or text datasets.
True/False?
Why is an RNN (Recurrent Neural Network) used for A-It can be trained as a supervised learning problem.
machine translation, say translating English to French?
(Check all that apply.) D-It is applicable when the input/output is a sequence (e.g., a sequence of words).
A-It can be trained as a supervised learning problem.
B-RNNs represent the recurrent process of Idea->Code-
>Experiment->Idea->....
C-It is strictly more powerful than a Convolutional Neural
Network (CNN).
D-It is applicable when the input/output is a sequence
(e.g., a sequence of words).
diagram: Performanc <- Amount of data True
Suppose the information given in the diagram is accurate. the graph shows that after a certain amount of data is fed to a NN it stops
We can deduce that when using large training sets, for a increasing its performance. To increase the performance it is necessary to use a
model to keep improving as the amount of data for training larger model.
grows, the size of the neural network must grow.
True/False?
, DPL-Course1-_Course4 Exam
diagram: Performanc <- Amount of data A-Increasing the training set size of a traditional learning algorithm stops helping to
improve the performance after a certain size.
Assuming the trends described in the figure are accurate.
Which of the following statements are true? Choose all that D-Increasing the size of a neural network generally does not hurt an algorithm's
apply. performance, and it may help significantly.
A-Increasing the training set size of a traditional learning
algorithm stops helping to improve the performance after a
certain size.
B-Decreasing the training set size generally does not hurt
an algorithm's performance, and it may help significantly.
C-Increasing the training set size of a traditional learning
algorithm always improves its performance.
D-Increasing the size of a neural network generally does
not hurt an algorithm's performance, and it may help
significantly.
**Which of these are reasons for Deep Learning recently B-Deep learning has resulted in significant improvements in important applications
taking off? (Check the three options that apply.) such as online advertising, speech recognition, and image recognition.
C-We have access to a lot more data. The digitalization of our society has played a
A-Neural Networks are a brand new field. huge role in this.
B-Deep learning has resulted in significant improvements D-We have access to a lot more computational power.
in important applications such as online advertising,
speech recognition, and image recognition.
C-We have access to a lot more data. The digitalization of
our society has played a huge role in this.
D-We have access to a lot more computational power.
**Recall this diagram of iterating over different ML ideas. A-Being able to try out ideas quickly allows deep learning engineers to iterate more
Which of the statements below are true? (Check all that quickly
apply.) B-Recent progress in deep learning algorithms has allowed us to train good models
faster (even without changing the CPU/GPU hardware). For example, we discussed
Idea->Code->Experiment->Idea how switching from sigmoid to ReLU activation functions allows faster training.
A-Being able to try out ideas quickly allows deep learning D-Faster computation can help speed up how long a team takes to iterate to a good
engineers to iterate more quickly idea.
B-Recent progress in deep learning algorithms has
allowed us to train good models faster (even without
changing the CPU/GPU hardware). For example, we
discussed how switching from sigmoid to ReLU activation
functions allows faster training.
C-It is faster to train on a big dataset than a small dataset.
D-Faster computation can help speed up how long a team
takes to iterate to a good idea.
Neural networks are good at figuring out functions relating True.
an input x to an output y given enough examples.
with neural networks, we don't need to "design" features by ourselves. The neural
True/False? network figures out the necessary relations given enough data.
**Which of the following are examples of unstructured B-Sound files for speech recognition.
data? Choose all that apply. C-Images for bird recognition.
A-Information about elephants' weight, height, age, and D-Text describing size and number of pages of books.
the number of offspring.
B-Sound files for speech recognition.
C-Images for bird recognition.
D-Text describing size and number of pages of books.
Which of the following are examples of structured data? B-A dataset with zip code, income, and name of a person.
Choose all that apply. C-A dataset of weight, height, age, the sugar level in the blood, and arterial
A-A set of audio recordings of a person saying a single pressure.
word.
B-A dataset with zip code, income, and name of a person.
C-A dataset of weight, height, age, the sugar level in the
blood, and arterial pressure.
D-A dataset with short poems.
, DPL-Course1-_Course4 Exam
Assuming the trends described in the figure are accurate. False.
The performance of a NN depends only on the size of the
NN. According to the trends in the figure above, It also depends on the amount of data.
True/False?
What does the analogy "AI is the new electricity" refer to? C-Similar to electricity starting about 100 years ago, AI is transforming multiple
A-AI runs on computers and is thus powered by electricity, industries.
but it is letting computers do things not possible before.
B-AI is powering personal devices in our homes and AI is transforming many fields from the car industry to agriculture to supply-chain...
offices, similar to electricity.
C-Similar to electricity starting about 100 years ago, AI is
transforming multiple industries.
D-Through the "smart grid", AI is delivering a new wave of
electricity.
When building a neural network to predict housing price False.
from features like size, the number of bedrooms, zip code,
and wealth, it is necessary to come up with other features A neural network figures out by itself the "features" in between using the samples
in between input and output like family size and school used to train it.
quality.
True/False?
Why can an RNN (Recurrent Neural Network) be used to A-The RNN is applicable since the input and output of the problem are sequences.
create English captions to French movies? Choose all that
apply. D-It can be trained as a supervised learning problem.
A-The RNN is applicable since the input and output of the
problem are sequences.
B-RNNs are much more powerful than a Convolutional
neural Network (CNN).
C-The RNN requires a small number of examples.
D-It can be trained as a supervised learning problem.
Which of the following are reasons that didn't allow Deep B-Limited computational power.
Learning to be developed during the '80s?
A-People were afraid of a machine rebellion. D-Interesting applications such as image recognition require large amounts of data
B-Limited computational power. that were not available.
C-The theoretical tools didn't exist during the 80's.
D-Interesting applications such as image recognition
require large amounts of data that were not available.
ReLU stands for which of the following? B-Rectified Linear Unit
A-Rectified Last Unit
B-Rectified Linear Unit
C-Recognition Linear Unit
D-Representation Linear Unit
diagram: Performanc <- Amount of data A-Increasing the training set size generally does not hurt an algorithm's
performance, and it may help significantly.
Assuming the trends described in the previous question's
figure are accurate (and hoping you got the axis labels C-Increasing the size of a neural network generally does not hurt an algorithm's
right), which of the following are true? (Check all that performance, and it may help significantly.
apply.)
A-Increasing the training set size generally does not hurt
an algorithm's performance, and it may help significantly.
B-Decreasing the training set size generally does not hurt
an algorithm's performance, and it may help significantly.
C-Increasing the size of a neural network generally does
not hurt an algorithm's performance, and it may help
significantly.
D-Decreasing the size of a neural network generally does
not hurt an algorithm's performance, and it may help
significantly.
, DPL-Course1-_Course4 Exam
RNNs (Recurrent Neural Networks) are good for data with True.
a temporal component.
RNN are good to work with sequences, and the elements of the sequence can be
True/False? sorted by a temporal component.
diagram: Performanc <- Amount of data False.
From the given diagram, we can deduce that Large NN when the amount of data is not large the performance of traditional learning
models are always better than traditional learning algorithms is shown to be the same as NN.
algorithms.
True/False?
Which of the following best describes the role of AI in the B-AI is an essential ingredient in realizing tasks, in industry and in personal life.
expression "an AI-powered society"?
In an AI-powered society AI plays a fundamental role to complete most tasks, in
A-AI controls the power grids for energy distribution, so all industry and personal life.
the power needed for industry and in daily life comes from
AI.
B-AI is an essential ingredient in realizing tasks, in industry
and in personal life.
C-AI helps to create a more efficient way of producing
energy to power industries and personal devices.
Features of animals, such as weight, height, and color, are TrueYes. The data can be represented by columns of data. This is an example of
used for classification between cats, dogs, or others. This structured data, unlike images of the animal.
is an example of "structured" data, because they are
represented as arrays in a computer. True/False?
FalseNo. The data can be represented by columns of data.
This is an example of structured data, unlike images of the
animal.
TrueYes. The data can be represented by columns of data.
This is an example of structured data, unlike images of the
animal.
A dataset is composed of age and weight data for several True
people. This dataset is an example of "structured" data
because it is represented as an array in a computer. Yes, the sequences can be represented as arrays in a computer. This is an example
of structured data.
True/False?
Recall this diagram of iterating over different ML ideas. B-Improvements in the GPU/CPU hardware enable the discovery of better Deep
Which of the statements below are true? (Check all that Learning algorithms.
apply.)
D-Better algorithms can speed up the iterative process by reducing the necessary
Graph: computation time.
Idea->Code->Experiment->Idea
A-Better algorithms allow engineers to get more data and
then produce better Deep Learning models.
B-Improvements in the GPU/CPU hardware enable the
discovery of better Deep Learning algorithms.
C-Larger amounts of data allow researchers to try more
ideas and then produce better algorithms in less time.
D-Better algorithms can speed up the iterative process by
reducing the necessary computation time.
What does a neuron compute? A-A neuron computes a linear function z=Wx+b followed by an activation function
A-A neuron computes a linear function z=Wx+b followed
by an activation function we generally say that the output of a neuron is a = g(Wx + b) where g is the
B-A neuron computes the mean of all features before activation function (sigmoid, tanh, ReLU, ...).
applying the output to an activation function
C-A neuron computes an activation function followed by a
linear function z=Wx+b
D-A neuron computes a function g that scales the input x
linearly (Wx + b)
Which of the following are some aspects in which AI has A-Web searching and advertisement.
transformed business?
AI has helped to make a fit between services or results and consumers or queries.
A-Web searching and advertisement.
B-Creating an AI-powered society.
C-Eliminating the need for health care services.
D-AI has not been able to transform businesses.
Which of the following play a major role to achieve a very A-Large amounts of data.
high level of performance with Deep Learning algorithms?
A-Large amounts of data. C-Large models
B-Better designed features to use. D-Deep learning has resulted in significant improvements in important applications
C-Large models. such as online advertising, speech recognition, and image recognition.
D-Deep learning has resulted in significant improvements
in important applications such as online advertising,
speech recognition, and image recognition.
E-Smaller models.
Recall the diagram of iterating over different ML ideas. A-Experiments finish faster, producing better ideas through increased iteration
Which of the stages shown in the diagram was improved tempo.
with the use of a better GPU/CPU?
C-Some algorithms are specifically designed to run experiments faster.
A-Experiments finish faster, producing better ideas through
increased iteration tempo.
B-Without better hardware, there is no way to train models
faster.
C-Some algorithms are specifically designed to run
experiments faster.
D-With larger datasets, the iteration process is faster.
When experienced deep learning engineers work on a new False
problem, they can usually use insight from previous
problems to train a good model on the first try, without Finding the characteristics of a model is key to having good performance. Although
needing to iterate multiple times through different models. experience can help, it requires multiple iterations to build a good model.
True/False?
Images for cat recognition is an example of "structured" False
data, because it is represented as a structured array in a
computer. Images for cat recognition are examples of "unstructured" data.
True/False?
A demographic dataset with statistics on different cities' False
population, GDP per capita, and economic growth is an
example of "unstructured" data because it contains data A demographic dataset with statistics on different cities' population, GDP per capita,
coming from different sources. and economic growth is an example of "structured" data in contrast to image, audio
or text datasets.
True/False?
Why is an RNN (Recurrent Neural Network) used for A-It can be trained as a supervised learning problem.
machine translation, say translating English to French?
(Check all that apply.) D-It is applicable when the input/output is a sequence (e.g., a sequence of words).
A-It can be trained as a supervised learning problem.
B-RNNs represent the recurrent process of Idea->Code-
>Experiment->Idea->....
C-It is strictly more powerful than a Convolutional Neural
Network (CNN).
D-It is applicable when the input/output is a sequence
(e.g., a sequence of words).
diagram: Performanc <- Amount of data True
Suppose the information given in the diagram is accurate. the graph shows that after a certain amount of data is fed to a NN it stops
We can deduce that when using large training sets, for a increasing its performance. To increase the performance it is necessary to use a
model to keep improving as the amount of data for training larger model.
grows, the size of the neural network must grow.
True/False?
, DPL-Course1-_Course4 Exam
diagram: Performanc <- Amount of data A-Increasing the training set size of a traditional learning algorithm stops helping to
improve the performance after a certain size.
Assuming the trends described in the figure are accurate.
Which of the following statements are true? Choose all that D-Increasing the size of a neural network generally does not hurt an algorithm's
apply. performance, and it may help significantly.
A-Increasing the training set size of a traditional learning
algorithm stops helping to improve the performance after a
certain size.
B-Decreasing the training set size generally does not hurt
an algorithm's performance, and it may help significantly.
C-Increasing the training set size of a traditional learning
algorithm always improves its performance.
D-Increasing the size of a neural network generally does
not hurt an algorithm's performance, and it may help
significantly.
**Which of these are reasons for Deep Learning recently B-Deep learning has resulted in significant improvements in important applications
taking off? (Check the three options that apply.) such as online advertising, speech recognition, and image recognition.
C-We have access to a lot more data. The digitalization of our society has played a
A-Neural Networks are a brand new field. huge role in this.
B-Deep learning has resulted in significant improvements D-We have access to a lot more computational power.
in important applications such as online advertising,
speech recognition, and image recognition.
C-We have access to a lot more data. The digitalization of
our society has played a huge role in this.
D-We have access to a lot more computational power.
**Recall this diagram of iterating over different ML ideas. A-Being able to try out ideas quickly allows deep learning engineers to iterate more
Which of the statements below are true? (Check all that quickly
apply.) B-Recent progress in deep learning algorithms has allowed us to train good models
faster (even without changing the CPU/GPU hardware). For example, we discussed
Idea->Code->Experiment->Idea how switching from sigmoid to ReLU activation functions allows faster training.
A-Being able to try out ideas quickly allows deep learning D-Faster computation can help speed up how long a team takes to iterate to a good
engineers to iterate more quickly idea.
B-Recent progress in deep learning algorithms has
allowed us to train good models faster (even without
changing the CPU/GPU hardware). For example, we
discussed how switching from sigmoid to ReLU activation
functions allows faster training.
C-It is faster to train on a big dataset than a small dataset.
D-Faster computation can help speed up how long a team
takes to iterate to a good idea.
Neural networks are good at figuring out functions relating True.
an input x to an output y given enough examples.
with neural networks, we don't need to "design" features by ourselves. The neural
True/False? network figures out the necessary relations given enough data.
**Which of the following are examples of unstructured B-Sound files for speech recognition.
data? Choose all that apply. C-Images for bird recognition.
A-Information about elephants' weight, height, age, and D-Text describing size and number of pages of books.
the number of offspring.
B-Sound files for speech recognition.
C-Images for bird recognition.
D-Text describing size and number of pages of books.
Which of the following are examples of structured data? B-A dataset with zip code, income, and name of a person.
Choose all that apply. C-A dataset of weight, height, age, the sugar level in the blood, and arterial
A-A set of audio recordings of a person saying a single pressure.
word.
B-A dataset with zip code, income, and name of a person.
C-A dataset of weight, height, age, the sugar level in the
blood, and arterial pressure.
D-A dataset with short poems.
, DPL-Course1-_Course4 Exam
Assuming the trends described in the figure are accurate. False.
The performance of a NN depends only on the size of the
NN. According to the trends in the figure above, It also depends on the amount of data.
True/False?
What does the analogy "AI is the new electricity" refer to? C-Similar to electricity starting about 100 years ago, AI is transforming multiple
A-AI runs on computers and is thus powered by electricity, industries.
but it is letting computers do things not possible before.
B-AI is powering personal devices in our homes and AI is transforming many fields from the car industry to agriculture to supply-chain...
offices, similar to electricity.
C-Similar to electricity starting about 100 years ago, AI is
transforming multiple industries.
D-Through the "smart grid", AI is delivering a new wave of
electricity.
When building a neural network to predict housing price False.
from features like size, the number of bedrooms, zip code,
and wealth, it is necessary to come up with other features A neural network figures out by itself the "features" in between using the samples
in between input and output like family size and school used to train it.
quality.
True/False?
Why can an RNN (Recurrent Neural Network) be used to A-The RNN is applicable since the input and output of the problem are sequences.
create English captions to French movies? Choose all that
apply. D-It can be trained as a supervised learning problem.
A-The RNN is applicable since the input and output of the
problem are sequences.
B-RNNs are much more powerful than a Convolutional
neural Network (CNN).
C-The RNN requires a small number of examples.
D-It can be trained as a supervised learning problem.
Which of the following are reasons that didn't allow Deep B-Limited computational power.
Learning to be developed during the '80s?
A-People were afraid of a machine rebellion. D-Interesting applications such as image recognition require large amounts of data
B-Limited computational power. that were not available.
C-The theoretical tools didn't exist during the 80's.
D-Interesting applications such as image recognition
require large amounts of data that were not available.
ReLU stands for which of the following? B-Rectified Linear Unit
A-Rectified Last Unit
B-Rectified Linear Unit
C-Recognition Linear Unit
D-Representation Linear Unit
diagram: Performanc <- Amount of data A-Increasing the training set size generally does not hurt an algorithm's
performance, and it may help significantly.
Assuming the trends described in the previous question's
figure are accurate (and hoping you got the axis labels C-Increasing the size of a neural network generally does not hurt an algorithm's
right), which of the following are true? (Check all that performance, and it may help significantly.
apply.)
A-Increasing the training set size generally does not hurt
an algorithm's performance, and it may help significantly.
B-Decreasing the training set size generally does not hurt
an algorithm's performance, and it may help significantly.
C-Increasing the size of a neural network generally does
not hurt an algorithm's performance, and it may help
significantly.
D-Decreasing the size of a neural network generally does
not hurt an algorithm's performance, and it may help
significantly.
, DPL-Course1-_Course4 Exam
RNNs (Recurrent Neural Networks) are good for data with True.
a temporal component.
RNN are good to work with sequences, and the elements of the sequence can be
True/False? sorted by a temporal component.
diagram: Performanc <- Amount of data False.
From the given diagram, we can deduce that Large NN when the amount of data is not large the performance of traditional learning
models are always better than traditional learning algorithms is shown to be the same as NN.
algorithms.
True/False?
Which of the following best describes the role of AI in the B-AI is an essential ingredient in realizing tasks, in industry and in personal life.
expression "an AI-powered society"?
In an AI-powered society AI plays a fundamental role to complete most tasks, in
A-AI controls the power grids for energy distribution, so all industry and personal life.
the power needed for industry and in daily life comes from
AI.
B-AI is an essential ingredient in realizing tasks, in industry
and in personal life.
C-AI helps to create a more efficient way of producing
energy to power industries and personal devices.
Features of animals, such as weight, height, and color, are TrueYes. The data can be represented by columns of data. This is an example of
used for classification between cats, dogs, or others. This structured data, unlike images of the animal.
is an example of "structured" data, because they are
represented as arrays in a computer. True/False?
FalseNo. The data can be represented by columns of data.
This is an example of structured data, unlike images of the
animal.
TrueYes. The data can be represented by columns of data.
This is an example of structured data, unlike images of the
animal.
A dataset is composed of age and weight data for several True
people. This dataset is an example of "structured" data
because it is represented as an array in a computer. Yes, the sequences can be represented as arrays in a computer. This is an example
of structured data.
True/False?
Recall this diagram of iterating over different ML ideas. B-Improvements in the GPU/CPU hardware enable the discovery of better Deep
Which of the statements below are true? (Check all that Learning algorithms.
apply.)
D-Better algorithms can speed up the iterative process by reducing the necessary
Graph: computation time.
Idea->Code->Experiment->Idea
A-Better algorithms allow engineers to get more data and
then produce better Deep Learning models.
B-Improvements in the GPU/CPU hardware enable the
discovery of better Deep Learning algorithms.
C-Larger amounts of data allow researchers to try more
ideas and then produce better algorithms in less time.
D-Better algorithms can speed up the iterative process by
reducing the necessary computation time.
What does a neuron compute? A-A neuron computes a linear function z=Wx+b followed by an activation function
A-A neuron computes a linear function z=Wx+b followed
by an activation function we generally say that the output of a neuron is a = g(Wx + b) where g is the
B-A neuron computes the mean of all features before activation function (sigmoid, tanh, ReLU, ...).
applying the output to an activation function
C-A neuron computes an activation function followed by a
linear function z=Wx+b
D-A neuron computes a function g that scales the input x
linearly (Wx + b)
