Neural Networks and Deep Learning (1)

Introduction to Deep Learning

Quiz

728x90

Q1

What does the analogy “AI is the new electricity” refer to?

1. AI is powering personal devices in our homes and offices, similar to electricity.
2. Like electricity started about 100 years ago, AI is transforming multiple industries.
3. Through the “smart grid”, AI is delivering a new wave of electricity.
4. AI runs on computers and is thus powered by electricity, but it is letting computers do things not possible before.

Answer

2

Yes. AI is transforming many fields from the car industry to agriculture to supply-chain…

Q2

Which of the following best describes the role of AI in the expression "an AI-powered society"?

1. AI helps to create a more efficient way of producing energy to power industries and personal devices.
2. AI is an essential ingredient in realizing tasks, in industry and personal life.
3. AI controls the power grids for energy distribution, so all the power needed for industry and daily life comes from AI.

Answer

2

In an AI-powered society, AI plays a fundamental role in completing most tasks, in industry and personal life.

Q3

Which reasons didn't allow Deep Learning to be developed during the '80s?

1. Limited computational power.
2. Interesting applications such as image recognition require large amounts of data that were not available.
3. People were afraid of a machine rebellion.
4. The theoretical tools didn’t exist during the 80’s.

Answer

1

Yes. Deep Learning methods need a lot of computational power, and only recently the use of GPUs has accelerated the experimentation with Deep Learning.

2

Yes. Many resources used today to train Deep Learning projects come from the fact that our society digitizes almost everything, creating a large dataset to train Deep Learning models.

Q4

Which of these are reasons for Deep Learning recently taking off? (Check the three options that apply.)

1. We have access to a lot more computational power.
2. Neural Networks are a brand new field.
3. We have access to a lot more data.
4. Deep learning has significantly improved important applications such as online advertising, speech recognition, and image recognition.

Answer

1

Yes! The development of hardware, perhaps especially GPU computing, has significantly improved deep learning algorithms' performance.

3

Yes! The digitalization of our society has played a huge role in this.

4

These were all examples discussed in lecture 3.

Q5

Which of the following are examples of unstructured data? Choose all that apply.

1. Sound files for speech recognition.
2. Text describing the size and number of pages of books.
3. Information about elephants’ weight, height, age, and the number of offspring.
4. Images for bird recognition.

Answer

1, 2, 4

Q6

Which of the following plays a major role in achieving a very high level of performance with Deep Learning algorithms?

1. Large models
2. Large amounts of data
3. Deep learning has significantly improved important applications such as online advertising, speech recognition, and image recognition.
4. Smaller models
5. Better-designed features to use

Hint

One main difference between "classical" machine learning algorithms and deep learning algorithms is that Deep Learning models "figure out" the best features using the hidden layers.

Answer

1, 2

Q7

Recall the diagram of iterating over different ML ideas. Which of the stages shown in the diagram was improved with the use of a better GPU/CPU?

1. With larger datasets, the iteration process is faster.
2. Some algorithms are specifically designed to run experiments faster.
3. Experiments finish faster, producing better ideas through increased iteration tempo.
4. Without better hardware, there is no way to train models faster.

Hint

Training with more data usually requires more time.

The creation of better algorithms can reduce the time needed to train a model. Recall the effect of introducing the ReLU function.

Answer

2

Yes. According to the trends in the figure above, It also depends on the amount of data.

3

Yes. The experiments help to test ideas, by getting the feedback from the experiments new variations can be tested and the results might indicate new directions to explore.

Q8

When building a neural network to predict housing prices from features like size, the number of bedrooms, zip code, and wealth, it is necessary to come up with other features in between input and output like family size and school quality. True/False?

 

Answer

False

A neural network figures out by itself the "features" in between using the samples used to train it.

Q9

When experienced deep learning engineers work on a new problem, they can usually use insight from previous problems to train a good model on the first try, without needing to iterate multiple times through different models. True/False?

 

Hint

Finding the characteristics of a model is key to having good performance. Although experience can help, it requires multiple iterations to build a good model.

Answer

False

Yes. Many resources used today to train Deep Learning projects come from the fact that our society digitizes almost everything, creating a large dataset to train Deep Learning models.

Q10

Features of animals, such as weight, height, and color, are used for classification between cats, dogs, or others. This is an example of "structured" data because they are represented as arrays in a computer. True/False?

1. False No. The data can be represented by columns of data. This is an example of structured data, unlike images of the animal.
2. True Yes. The data can be represented by columns of data. This is an example of structured data, unlike images of the animal.

Answer

2

Q11

A demographic dataset with statistics on different cities' populations, GDP per capita, and economic growth is an example of “unstructured” data because it contains data coming from different sources. True/False?

 

Answer

False

A demographic dataset with statistics on different cities' populations, GDP per capita, and economic growth is an example of “structured” data in contrast to image, audio, or text datasets.

Q12

Why can an RNN (Recurrent Neural Network) be used to create English captions for French movies? Choose all that apply.

1. RNNs are much more powerful than a Convolutional neural Network (CNN).
2. It can be trained as a supervised learning problem.
3. The RNN is applicable since the input and output of the problem are sequences.
4. The RNN requires a small number of examples.

Answer

2

The data can be used as x (movie audio) to y (caption text).

3

Yes, an RNN can map from a sequence of sounds (or audio files) to a sequence of words (the caption).

Q13

From the diagram given, we can deduce that large NN models are always better than traditional learning algorithms. True/False?

Answer

False

Yes, when the amount of data is not large the performance of traditional learning algorithms is shown to be the same as NN.

Q14

Assuming the trends described in the figure are accurate. The performance of a NN depends only on the size of the NN. True/False?

Answer

False

Yes. According to the trends in the figure above, It also depends on the amount of data.

All the information provided is based on the Deep Learning Specialization | Coursera from DeepLearning.AI