Skip to main content
. Author manuscript; available in PMC: 2009 Jun 25.
Published in final edited form as: Neurocomputing (Amst). 2007 Aug 1;70(13-15):2245–2255. doi: 10.1016/j.neucom.2006.03.019

Fig. 8.

Fig. 8

Comparison of output errors. The mean error in the output vs. the target value in each trial and its 99% confidence interval (error bar) are shown for all test cases (the plotting convention is the same as Figure 7). In all cases, the differences between the 3D-net and the 2D-net were significant (t-test: n = 4, 800, p < 0.02), except for S2 2D (p = 0.47) where the performance was comparable. Note that for the 2D representation of S1, 2D-net < 3D-net.