Dream Simulation (Periodic)

Posted: November 1, 2013 at 4:48 pm

I ran a test training the MLP with data filtered such that at least one cluster must change state between subsequent frames. This time, the result is a periodic dream that does not stabilize over 10,000 iterations:

backgroundState-filtered-1.dream-crop

On a hunch I figured the network state would be highly dependent on initial conditions, which is true. If I feed the exact same network as above with the last data-point in the set (rather than the first data-point as in previous dreams), we get a static result after a single iteration:

backgroundState-filtered-1-test.dream-crop