Deck 6: Neural Networks

Question 1

Compare and contrast linear relationships and non-linear relationships.

Answer 1

No Answer

Question 2

List and explain the key reminders to be used when using neural networks.

Answer 2

No Answer

Question 3

In the context of activation functions, which of the statements is true?

A) Activation functions are used to identify linear regressions.
B) Weights decrease the speed in switching on the activation functions.
C) Bias in the network speeds up the activation functions.
D) Activation functions help the model to detect non-linear relationships.

Answer 3

Activation functions help the model to detect non-linear relationships.

Question 4

Explain the three layers of a neural network and the functions of each layer.

Question 5

Identify a true statement about the hidden layer of a neural network.

A) Calculations are carried out in this method and weights are produced from the input layer.
B) This layer is used to input data into the neural network for analysis.
C) The neural network model arrives at a prediction in the hidden layer.
D) Data from the output layer is transferred to the hidden layer.

Question 6

A health insurance company uses Neuralmind to drive its customers to log on to its website and complete a primary health assessment and an insurance quote. This is an example of using a neural network to

A) analyze audience sentiment.
B) entice customers to a brick-and-mortar store.
C) predict lead scoring.
D) increase new customers.

Question 7

In a neural network, the learning rate consists of negative values that generally range between -1 and -10.

Question 8

Summarize practitioner Stephen Brobst's thoughts on the uses and advantages of neural networks.

Question 9

In a neural network, weights are values representing features from the dataset that pass information to the next layer via connections.

Question 10

Which of the following is true of training a neural network?

A) A neural network will stop training itself when it reaches a particular threshold or maximum run limits.
B) If a neural network is trained too much, the result is underfitting leading to underprediction.
C) A neural network is considered to be optimally trained when model prediction no longer improves.
D) If a neural network is not trained enough, the result will be overfitting of the model.

Question 11

In a neural network, inputs that are important in predicting the output have smaller weights, whereas the less important inputs have larger weights.

Question 12

An advertising company uses a neural network software to determine which buyers are most likely to open their promotional emails based on past purchase behavior. Using data from 20 previous email campaigns, the neural network trains itself to examine the impact of 25 features and develop recommended solutions. The results almost doubled customers' response rates to 8.2 percent, which in turn, reduced product promotion costs by 35 percent. This example illustrates the use of neural network technology to

A) increase new customers that often request online product quotes.
B) determine customer lifetime value and well-defined customer segments.
C) develop new products or make product and service recommendations to customers.
D) classify customers by their likely profitability when planning direct marketing strategies.

Question 13

According to neural networks practitioner Stephen Brobst, a neural network technology is most likely to be used when the dataset is

A) clean.
B) highly dimensional.
C) complete.
D) smaller than what is required for linear regression.

Question 14

In a neural network, the learning rate determines the speed at which the model can reach the most accurate solution.

Question 15

Which of the following is an example of using a neural network to create products and make service recommendations?

A) Cricket Wireless works with Cognitiv to come up with solutions to predict the likelihood of non-Cricket customers visiting the store and to develop digital advertising campaigns.
B) Microsoft uses BrainMaker, a neural network software, to determine which customers are most likely to open their direct mail based on past purchase behavior.
C) Netflix uses neural networks to develop insights into viewer preferences to improve production and procurement of relevant movies.
D) ASOS uses neural networks to analyze customer behavior that occurs on the website to predict the value of a customer.

Question 16

Describe how marketing strategists employ neural networks to answer business problems and make decisions.

Question 17

Regression is useful for forecasting consumer behavior when predictive relationships are linear.

Question 18

Within a neural network, nodes or neurons

A) are values that represent features from the dataset.
B) determine the amount of adjustment made to the weights in the network.
C) determine the constant value given to the weighted input of each feature.
D) are a set of inputs that are multiplied by their weights.

Question 19

According to practitioner Stephen Brobst, which of the following is the best approach for cost-effective predictive capabilities?

A) to rely heavily on deep learning and neural networks for maximum predictability
B) to retain and use traditional machine learning models as they are much cheaper to execute
C) to think of traditional machine learning and deep learning as complementary
D) to find that one best algorithm that predicts customer behavior most precisely

Question 20

Illustrate with a short example and description how a neural network learns.

Question 21

In a neural network, which of the following occurs during the process of forward propagation?

A) determination of the speed at which the model can arrive at the most accurate solution
B) sending of data through the hidden layer where it is processed using an activation function
C) transmission of total loss back into the neural network to understand the amount of loss from each neuron
D) determination of the amount of adjustment made to the weights in the network

Question 22

The Proctor & Gamble brand Olay has a mobile application that enables customers to obtain skincare assessments. The app examines the customer's image to determine potential skin issues and then recommends specific products to address areas of concern. This is an example of using a neural network to

A) assess market sentiment.
B) determine the value of a potential buyer.
C) personalize customer experiences.
D) predict lead scoring.

Question 23

A neural network is constantly learning until the final optimized solution is obtained, which is when model prediction is no longer improving.

Question 24

Compare and contrast forward propagation and backward propagation.

Question 25

Explain how a digital content streaming platform such as Netflix might employ a neural network to improve business.

Question 26

In a neural network, inputs

A) are variables from the dataset that move information to the next layer via connections.
B) are the constant value given to the weighted input of each node.
C) determine the speed at which the model can arrive at the most accurate solution.
D) determine the amount of adjustment made to the weights in the network.

Question 27

The logic of neural networks is based on patterns recognized by observing biological activities in the human brain.

Question 28

DialogTech provides neural network-driven marketing analytics solutions to manage customer inbound call centers. The collected data includes incoming caller objectives, interactions between the callers and salespersons, and assessment of conversation outcomes. This example from the text illustrates the use of neural network technology to

A) generate personalized content.
B) categorize customers.
C) predict lead scoring.
D) build a dataset.

Question 29

Neural networks

A) often do not work well with out-of-range data.
B) have simple architectures.
C) require small datasets for optimum prediction.
D) do not perform properly when only key predictors are included.

Question 30

HealthX, a fitness products company, launched a mobile application that enables customers to obtain fitness assessments. The app analyzes customers' data to determine potential health issues and then recommends specific measures and products to address areas of concern. The digital interaction mimics an in-person transaction while allowing customers to remain in the comfort of their home. This example illustrates the use of neural network technology to

A) analyze audience sentiment to improve a product.
B) entice customers to visit a brick-and-mortar store.
C) personalize customer experiences.
D) predict lead scoring.

Question 31

In a neural network, the output is calculated as

A) Output = inputs (bias × sum) + weights.
B) Output = bias (weights × inputs) + sum.
C) Output = weights (sum × bias) + inputs.
D) Output = sum (weights × inputs) + bias.

Question 32

Under Armour has a health-tracking mobile application known as Record. The app collects health-related data from a variety of sources, such as manually entered user data, wearable devices, and other third-party applications. The data includes sleeping patterns, workouts, nutrition, and related information that can best be used in a neural network to

A) develop customized digital content, such as exercise and diet recommendations for its app users.
B) entice customers to visit a brick-and-mortar store.
C) classify new customers by their potential profitability when planning direct marketing strategies.
D) predict a customer lead score.

Question 33

Shield, a fitness products company, has a health-tracking mobile application called GoFit. The app collects health-related data from sources such as wearable devices and manually entered user data. The data includes sleeping patterns, workouts, nutrition, and related information. Shield can best use this information in a neural network to

A) entice customers to visit a brick-and-mortar store.
B) predict lead scoring.
C) develop customized digital content, such as exercise and diet recommendations for its app users.
D) classify new customers by their potential profitability when planning direct marketing strategies.

Question 34

Which of the following is true of neural networks?

A) They are algorithms trained to recognize patterns in large volumes of data.
B) They are cheaper and simpler to execute as compared to linear regression methods.
C) They are based on the chemical reactions that make up brain signals.
D) They are most useful when predictive relationships are linear.

Question 35

A marketing company uses BrainMaker to identify the customers who are most likely to click on their direct mail based on their past purchase behavior. This resulted in a 35 percent drop in advertising costs. This is an example of using a neural network to

A) analyze market sentiment to alter advertising approaches.
B) innovate and design new products and categories.
C) classify new customers by their likely profitability when planning direct marketing strategies.
D) entice customers to visit a brick-and-mortar store.

Question 36

Neural networks can aid marketers in predicting customer behavior, understanding buyer segmentation, developing brand strategies, forecasting sales, optimizing inventory, improving marketing automation, developing digital content, and much more.

Question 37

Neural networks are acknowledged for their good predictive performance and can be generalized to new values.

Question 38

An automotive insurance company hired Cognitiv to apply their deep learning platform, Neuralmind, to drive prospects to log on to their website and complete an insurance quote online. This is an example of using a neural network to

A) analyze audience sentiment.
B) increase new customers.
C) entice customers to a brick-and-mortar store.
D) predict lead scoring.

Question 39

In a neural network, the hidden layer sits on top of the input and output layers.