| |
Results
|
The following
picture
represents a part of the prototypes (a prototype of a neuron is inside a
square) of a 8000 neurons map. Topological property of self organizing map
is preserved in our model. However, locally, variations of contrast aren't
exhaustive. |
We have two manners to explain this lack of exhaustivity :
| the world where the robot evolves isn't rich enough to express the whole
possible variations of contrast. |
| the property of retinotopy implies strong constraints during the self
organization process. In order to solve this problem, we have modified the
inputs of the cortical map in order to free the process. Our solution ,
however, doesn't free the process enough. |
Our model extends the Kohonen map by adding it the retinotopy. In order
to free the process of self organisation, we need to look for a model that takes
into account biological properties more accurately.
|