Receptive Field
A portion of sensory space that can elicit neuronal responses (in the [see page 5, visual] pathway) when stimulated. Essentially when we have a screen with some stimulus (light) being shone on it then certain neurons will respond quite strongly when stimuli are in certain positions or coming from certain directions.
In the Visual Pathway
Receptive fields can be devided into categories depending on how deep into the visual pathway their found.
Retina (Lateral Geniculate Nucleus)
Very simple fields found earlier into the pathway and have basic excitory or inhibitory behavior resembling a simple two-layer neural network (N-input -> 1-output).
These can be [see page 8, either]:
| Term | Description |
|---|---|
| On-center | Excitory at the center of the field, but inhibitory in some range leading up to it. |
| Off-center | Inhibitory at the center, but excitory in some range leading up to it. |
Visual Cortex (V1)
Found deeper into the pathway and can demonstrate more specific behavior as a consequence of a more complex structure (due to complex wiring of the earlier Retinal structure, combining multiple on-center and off-center receptive fields in a form similar to a neural network).
These fields are [see page 9, orientation selective], they can be excitory or inhibatory at multiple points within the field and the [see page 11, direction] through which a stimulus is directed can impact the strength of the reflex (positives cancel out negatives and vice-versa).
Note: This requires very precise wiring, most likely through hebbian learning.
Hierarchy
The increasing complexity of receptive fields as we progress up the visual pathway can be modeled as a [see page 8, hierarchy]. From bottom to top we have:
- simple-cells/input-neurons (retina fields)
- complex cells combining retinal cells (example: detect lines)
- composite feature cells (example: different lines appearing together)
- complex composite cells
- view tuned cells (example: detect complete, joined paths)