Exam 3

Question 1

From the matlab training video, what are 2 examples of deep learning applications

Answer 1

• radiology reviews, preliminary diagonoses
• self driving algorthms

Question 2

Deep learning performs…

Answer 2

end to end learning. It receives raw data and learns to perform the task defined by itself.

Question 3

Shallow learning is…

Answer 3

supervised learning using an ANN with low complexity

Question 4

What distinguishes deep learning from shallow learning?

Answer 4

• higher complexity (will have multiple networks within a network, numerous hidden layers)
• input information is also more complex

Question 5

A common application of deep learning is…

Answer 5

image recognition

Question 6

DCNN stands for

Answer 6

Deep Convolution Neural Network

Question 7

DCNNs are commonly used for

Answer 7

image recognition and audio processing (computer vision!)

Question 8

What is a convolution layer?

Answer 8

A thin layer that extracts a reduced set of information, similar to filtering in signal processing

Question 9

A layer which takes detailed raw data you have an performs feature engineering is an example of a

Answer 9

convolution layer

Question 10

A pooling layer performs

Answer 10

dimensional reduction

Question 11

A fully connected layer…

Answer 11

restores global connections

Question 12

A dropout layer can help…

Answer 12

avoid overfitting and slowness

Question 13

What does a dropout layer do?

Answer 13

Randomly drops neurons and their connections in the network.

Simplifies the NN, analogous to pruning trees

Question 14

What is a deep feed-forward layer?

Answer 14

All linear left to right flow of data and calculations. No circular data or feedback loops.

Question 15

What can feedback loops be particularly useful for?

Answer 15

Helping the neural network learn from time dependent information.

Question 16

What are some features of a recurrent neural network?

Answer 16

• has circular connections
• good for dynamic and time dependent data

Question 17

What makes a network a “deep convolution” NN?

Answer 17

It has at least one or more convolution hidden layers

Question 18

A _____ performs the opposite action of a convolution neural network

Answer 18

a Deconvolution NN

Question 19

What is a Generative Adversarial Network?

Answer 19

• Generative = generates new content (images, texts, audio)
• Combination of DCNNs and FF

Question 20

What is Transfer Learning?

Answer 20

Using an already trained neural network for a new NN and a similar data set

Question 21

When performing transfer learning, you add…

Answer 21

Add a few new hidden layers on the right, and train on only those new layesr

Question 22

What is a tensor:

Answer 22

a higher dimensional matrix

Question 23

A _____ NN is used to classify and reduce dimensionality

Answer 23

Auto Encoder
classify = encode

Question 24

Shallow learning tends to have ___ hidden layers

Answer 24

one or two

Question 25

Deep learning is unsupervised or supervised

Answer 25

Supervised

Question 26

A color photo with 1000x1000 pixels would make what dimension matrix?

Answer 26

1000x1000x3
3 for the colors (assuming RGB combo)

Question 27

Gradient descent optimization gets very complex with DNNS, greater chance of getting stuck in a local minimum because there are….

Answer 27

Many many more weight terms to solve for (could be millions!)

Question 28

One approach to improve gradient based optimization for deep NNs is a…

Answer 28

stochastic approach (mentioned in lecture)

Question 29

A convolution layer extracts…

Answer 29

A subset of important information out of all the data you fed it (filtering)

Question 30

One practical purpose of a sequence of convolution and pooling layers is to…

Answer 30

reduce the information down to the important inputs needed to actually run the calculations in subsequent layers that have the main learning algorithm

Question 31

Recurrent NNs take more or less time to train than a typical FF NN?

Answer 31

More