Skip to main content
. Author manuscript; available in PMC: 2009 Apr 2.
Published in final edited form as: Brain Res. 2007 Jul 26;1202:14–24. doi: 10.1016/j.brainres.2007.06.079

Figure 9.

Figure 9

Time to learn Greebles over Greeble pre-training activation variance. As the variance of the hidden layer activations over the Greeble stimuli increases, the training required to learn Greebles decreases. This correlation is strong (r= -0.6317, p < 0.0001). This variance is taken before the networks are trained with Greebles and represents the initial spread of Greebles in representational space.