X hits on this document

PDF document

The split-fovea model of visual word - page 57 / 78

320 views

0 shares

0 downloads

0 comments

57 / 78

Chapter 2. The Simulations

52

Figure 2.8: Control net: MSE at every fixation position at epochs 2000-2750

Figure 2.9: Fixation net: MSE at every fixation position at epochs 2000-2750

Note that similar graphs can be drawn for the number of errors made by the networks at the different fixation positions. These graphs would show a similar pictures as the MSE graphs, with the positions having a low MSE also making a low number of errors and vice-versa. To conclude this section it can be said that these networks confirm the findings of the previous networks. There is no noticeable difference in the overall performance of the networks with both getting similar percentages of words correct. However, there is a difference in the more detailed analysis of the networks, namely the performance at the different fixation positions.

Document info
Document views320
Page views334
Page last viewedSun Jan 22 06:30:20 UTC 2017
Pages78
Paragraphs3087
Words17414

Comments