BI_Ch_07_questions

What is the main goal of deep learning in a business setting?

• To replace all databases

• To improve prediction and pattern recognition from data

• To eliminate the need for managers

• To reduce the use of information systems

Deep learning is best described as:

• A type of spreadsheet software

• A subset of machine learning that uses multilayer neural networks

• A manual forecasting method

• A rule-based accounting system

One important advantage of deep learning over many traditional machine-learning methods is:

• It always requires less data

• It automatically extracts useful features from raw data

• It does not need training

• It only works with financial statements

In an artificial neural network, a neuron mainly performs which task?

• It stores paper records

• It processes inputs and produces an output

• It deletes missing data

• It creates dashboards automatically

In a neural network, connection weights are used to:

• Measure office productivity

• Control the importance of input signals

• Replace all output values

• Eliminate hidden layers

Which of the following is a common activation function in neural networks?

• Log-sigmoid

• Bar chart

• Pivot table

• Decision table

What is the purpose of the hidden layer in a neural network?

• To store backup files

• To help detect patterns and relationships in the data

• To replace the input data

• To generate random outputs

During neural network training, the system improves its performance mainly by:

• Increasing the number of managers

• Adjusting the weights after comparing output with target values

• Removing all input variables

• Converting all data into text

What is backpropagation?

• A method for designing dashboards

• A training method that updates neural-network weights by working backward from the error

• A method for cleaning customer records

• A way to compress image files

Overfitting occurs when a neural network:

• Performs well on training data but poorly on new data

• Refuses to learn from any data

• Uses too few variables

• Produces the same output for all cases