Skip to main content
. 2008 Jan 15;98(3):259–272. doi: 10.1007/s00422-007-0209-6

Fig. 4.

Fig. 4

Comparison of two modified learning algorithms. a ICO-learning. b Acting-TD. While Acting-TD and all architectures shown in Fig. 2 use the derivative of the postsynaptic signal, ICO-learning a uses the derivative of the unconditioned signal in order to control weight change. Note, x 0 in Acting-TD is identical to the delta error δ