NEW
Font size
WorksheetsANN Part 4 - Part 6 Quiz
Total questions: 20
Worksheet time: 10mins
Which of the following is not a way to generate dataset for ANN?
Interview or survey
Random guessing
Hardware simulation
Previous performance records
In ANN datasets, what do the rows typically represent?
Output classes
Epochs
Input variables
Training accuracy
For a dataset with 70% training and 30% testing, how many testing patterns are there in a 10-pattern dataset?
7
3
5
2
What is the purpose of using a separate testing dataset in ANN?
To reduce cost
To check model generalisation
To increase speed
To avoid using historical data
Which Python command is used to normalise input data in ANN?
fit_transform()
scale_input()
sort_data()
mean_adjust()
What does denormalisation in ANN do?
Removes noise from the data
Reverts predicted values back to original scale
Removes outliers
Combines datasets
What type of value does the output represent in a classification problem?
Continuous
Category
Decimal
Boolean only
What are the two sequential phases in supervised ANN training?
Backpropagation and testing
Prediction and adjustment
Forwarding and adjustment
Training and saving
What does backpropagation calculate in ANN training?
Input size
Gradients of error
Memory usage
Number of neurons
Which of the following best describes the learning rate (α)?
Number of hidden layers
Rate of data input
Size of weight adjustment during training
Bias value update
What happens if the learning rate α is too high?
Model trains faster with perfect accuracy
Model may overshoot optimal solution
Model will stop learning
Model will underfit
What is the main role of the optimizer in ANN training?
Compute gradients
Choose number of layers
Apply gradient updates to weights
Set bias values
What is the role of momentum (β) in ANN training?
Avoid gradient vanishing
Prevent overfitting
Avoid local minima trap
Increase learning rate
Which is NOT a type of optimizer used in ANN?
ADAM
SGD
ReLU
RMSprop
Which command in Keras sets the optimizer and loss function?
model.add()
model.fit()
model.compile()
model.train()
What is the purpose of ANN testing?
To increase training time
To simulate unseen data
To reset weights
To normalise inputs
What does the correlation coefficient (r) measure in ANN testing?
Speed of convergence
Difference between two datasets
Relationship between predicted and actual output
Number of output classes
What is the desired value of correlation coefficient (r) for successful ANN testing?
r < 0
r = 0
r ≥ 0.9
r ≤ 0.5
Which equation describes the regression line in ANN testing?
Y = aX + b
Y = mX + c
Y = X + 1
Y = wX + b
What should you do if your correlation coefficient is below 0.9?
Accept the result
Reduce input variables
Retrain the network
Change the testing data
