Quiz 8 (part 2)

Question 1

True or false: Neural network weights should be initialised to large numbers

Question 2

True or false: An artificial neural network can work with missing values

Question 3

True or false: An artificial neural network could be trained to solve the logical OR problem

Question 4

True or false: Leaky rectified linear units is a type of activation function

Question 5

True or false: It is a good idea to scale the features before training a neural network

Question 6

True or false: Activation functions are only applied to the product of the weights and the features, and excludes the bias.

Question 7

Select one or more: With reference to error-based learning, the main purpose of the learning rate is to

• a. prevent the search direction from changing when stochastic learning is used.
• b. adjust the step sizes added to current weight values
• c. prevent the neural network to become stuck in a local minimum