Click on the image to see a PDF version (for zooming in)
Fig. 6.3. Architecture of the reduced LISSOM model. As long as
Gaussian patterns are used as input, it is not necessary to include
separate ON and OFF channels in the model. Instead, the input can be
directly presented as ON channel activations. For simplicity and
compatibility with other similar models, this channel is called the
Retina, and its neurons are referred to as photoreceptors. The
afferent connections form a local anatomical receptive field directly
on the redefined retina; the lateral connections are similar to those
in other LISSOM models. The reduced model is more efficient to
simulate computationally than the full LISSOM, with equivalent results
(as shown in Figures 6.4 and 6.5).
|