Another comparison was made on the ability to recognize noisy inputs. The noise on the pictures was generated randomly. The noise-level is the probability for each pixel to be altered to a random value in the interval [0,1] (an uniform distribution was used). In Figure 7.8 pictures with different noise-levels are shown.
Figure 7.8: Examples of Noisy Test Pictures.
The modular network could recognize pictures with a significant higher noise-level than the single MLP; the results are shown in Figure 7.9.
Figure 7.9: The Performance on Noisy Inputs.
From the above experiments it can be seen that the modular network has superior generalization abilities on this type of high dimensional input vectors.