COMP9444 Neural Networks

Exercises on Hopfield Networks

  1. Can the vector [1, 0, –1, 0, 1] be stored in a 5-neuron discrete Hopfield network? If so, what would be the weight matrix for a Hopfield network with just that vector stored in it? If not, why not?

    1. Compute the weight matrix for a Hopfield network with the two memory vectors [1, –1, 1, –1, 1, 1] and [1, 1, 1, –1, –1, –1] stored in it.
    2. Confirm that both these vectors are stable states of the network.

  2. Consider the following weight matrix W:
    0.0 –0.2 0.2 –0.2 –0.2
    –0.2 0.0 –0.2 0.2 0.2
    0.2 –0.2 0.0 –0.2 –0.2
    –0.2 0.2 –0.2 0.0 0.2
    –0.2 0.2 –0.2 0.2 0.0

    1. Starting in the state [1, 1, 1, 1, –1], compute the state flow to the stable state using asynchronous updates.
    2. Starting in the (same) state [1, 1, 1, 1, –1], compute the next state using synchronous updates.


Copyright © Bill Wilson, 2008.
Bill Wilson's contact info

UNSW's CRICOS Provider No. is 00098G