Read the summary and the most important questions on Slides

• 3 Neural Networks: Unsupervised Learning

  This is a preview. There are 17 more flashcards available for chapter 3
  Show more cards here

• In what two ways does a neural network resemble the brain?

  - Knowledge is acquired by the network from its environment through a learning  process.
  - Inter-neuron connection strengths, known as synaptic weights are used to store the acquired knowledge.

• In what way is a neural network a black box?

  We do not actually know what the network is learning.

• What are examples of neural networks?

  Perceptron, MLP, RBF.

• What does the basic model of a neuron look like?

  Three inputs, all with their own weight, and a bias (1) with weight number zero. Then the output a will be weight zero plus the sum of the input times its weight.

• What is the function of unsupervised learning?

  Discover significant patterns or features in data through the use of unlabelled examples.

• What are two different perspectives of unsupervised learning?

  - Self-organised learning: Algorithm gets rules of local behaviour that are used to compute an input-output mapping with certain desirable properties.
  - Statistical learning theory: More emphasis on mathematical tools.

• How would a PCA statistical learning approach be build up?

  1. Start with a dataset.
  2. Calculate the covariance matrix.
  3. Calculate the eigenvalue and eigenvectors.
  4. Transform the dataset using the eigenvectors with the highest values.

• PCA is possible through self-organisation. What are the principles of self-organisation?

  1. Self-amplification.
  2. Competition.
  3. Cooperation.
  4. Structural information.

• What is the first principle of self-organisation?

  Self-amplification: Synaptic weights self-amplify in accordance with Hebb's postulate of learning:
  - If two neurons on either side of a connection are activated simultaneously (i.e. synchronously), then the strength of that connection is selectively increased.
  - In the case of asynchronous activation the connection strength is decreased.

• What is the second principle of self-organisation?

  Competition: The availability of limited resources in one form or the other leads to competition among neurons with the result that the most vigorously growing neurons are selefcted at the expense of the others.
  - Neurons compete with each other in a winner takes all fashion.