COMP9844 Extended Neural Networks

Dynamic Nets - Solutions

It is best not to read the answers until you've tried to answer the questions yourself.

  1. What does it mean to say that a radial basis function has a centre and a radius?

    Answer: This means that, for a particular type f of RBF, the value of that the function takes depends only on the distance of the input vector from the "centre" vector, together with the radius parameter. Specifically, if the centre is c and the radius is r, the value taken for input vector v will be f(||cv||/r)

  2. What is the architecture of a radial basis function network?

    Answer: Three layers:
    Layer 1: input layer.
    Layer 2: RBF layer. Nodes compute an RBF using the inputs, and output it to layer 3.
    Layer 3: Weighted connections layer. Nodes compute the total net input (using the weights) from layer 2, and either output this linear combination, or perhaps transform it.

  3. What are the two stages in training a radial basis function network?

    Answer: stage 1: establish the centres and radii for the RBF layer. Typically an unsupervised learning algorithm is used.
    stage 2: discover the weights for the output layer. Typically a supervised learning algorithm is used for this stage.


Copyright © Bill Wilson, 2009.
Bill Wilson's contact info

UNSW's CRICOS Provider No. is 00098G