Self-assessment questions

  • In the context of neural networks, explain the terms: 
  1. artificial neuron, 
  2. activation function, 
  3. backpropagation, 
  4. supervised  training, 
  5. separability, 
  6. non-linear separable problem, 
  7. resilient propagation, 
  8. momentum term.
  • What are neural computing assumptions about computation in the brain? Explain.
  • Construct a multilayer network with four input nodes, three nodes in the middle layer and two output nodes. What is the role of this middle layer, why is this called a “hidden layer”, and what does it hide?
  • Describe the steps of the standard backpropagation training algorithm.
  • Explain the role of a decision boundary in the context of neural computing.
  • Minimising an error function often carries out neural network training. Describe a suitable error function for a feedforward neural network.
  • Discuss modifications of the BP algorithm. What kind of improvements can these modification offer?

Last modified: Friday, 7 February 2020, 12:37 PM