The first week of the "deeplearning.ai-Neural network and deep learning" answer

the first week after-school assignment is a 10-course choice question

Note: The answer is from the first one and then the ABCD ... The answer has its own understanding, there are also from the online blog reference, only to learn.
1. First question

I understand the answer: D.
Reference answer: A. "AI is the new power", this is the topic of Wunda Teacher's speech on AI conference this year. Of course, the analogy is that AI, like electricity 100 years ago, is bringing great changes to our productive lives. In a speech, Wunda said, "In the future, build an AI-driven society, and everything around us has AI intelligence and changes human life." "The realization of this dream is not only based on the efforts of a company, but the efforts of all of us."
2. The second question

Answer: ABD. The question asks us what are the reasons for the recent deep study of this fire. In other words, deep learning is not a new field, it is only one of the many machine learning algorithms, why in the ups and downs after the rise and fall behind, the recent can fire up. Because deep learning algorithms require a lot of data and strong hardware computing power. Previously limited by data volume and computing power, has been tepid. In recent years the Internet has flourished, all kinds of information have been realized data, the amount of data is greatly increased, you think of your online shopping when you stay on the Internet information you know. In addition, the computer hardware in accordance with the "Moore's Law" development, the exponential growth of computing power, which provides a good basis for the re-blooming of deep learning algorithms. Of course, can not be separated from the use of deep learning algorithms to achieve excellent performance of various products, such as alpha Dog, image recognition robot, Sophia Robot.

3. Third question

Answer: ABD. This diagram shows us a process of developing a neural network, from having ideas, to implementing them in code, to running the code to see the results, and then going back to revising the idea so that it repeats itself. If a team can implement ideas faster in code, or if the Code (algorithm) is better, or if scientists work out better algorithms, the cycle will be shorter and more efficient. For these reasons, training a large data set may not take longer than training a small data set, and of course it is not faster to train on big data sets.

4. Question Fourth

Answer: B. This question examines our understanding of operational experience and models. A well-behaved deep learning model is not something that can be found immediately by experience. Although experience is very important, to find a well-behaved model, you need to try, repair, and constantly improve a process.

5. Question Fifth

Answer: C. Common activation functions are: The sigmoid function, the Tanh function, and the leaky Relu function. The D option is the leaky Relu function.

6. Question Sixth

Answer: B. This topic examines our understanding of structured and unstructured data. Cat image recognition data is typical unstructured data, common unstructured data and text, images, video and so on.

7. Question Seventh

Answer: B. This topic examines our understanding of structured and unstructured data. Numbers such as population, GDP growth, and economic growth are structured data

8. Question Eighth

Answer: AC. This question examines our understanding of RNN (recurrent neural networks). RNN has achieved some success in speech recognition, language modeling, translation, picture description and other issues. It is a supervised learning, such as input data in English, labeled French. RNN can be seen as multiple assignments of the same neural network, and each neural network module transmits the message to the next, so it's chained, and the chain features reveal that RNN is inherently related to sequences and lists, so it's no problem in solving sequence. To say which is stronger than the other, is basically wrong.

9. Question Nineth

Answer: B. Deep learning, larger data matches a larger depth learning model, and the upper limit of its effect is currently not a bottleneck.

10. Question Tenth

Answer: BD. In general, for the same amount of data, as long as enough, large neural networks perform better. The more data you have on the same neural network, the better it behaves.