Orientation Demo

Spinning line and cortical response Orientation key Orientation map
a b c d e
g h
  1. retinal image presented to map g.
  2. initial response of the cortex to that image
  3. orientation histogram of the initial response
  4. settled response of the cortex to that image
  5. orientation histogram of the settled response
  6. key showing color assigned to each angle
  7. self-organized RF-LISSOM orientation map
  8. orientation histogram of the orientation map
To make this data easy to understand, all learning and motion were disabled in the model. Thus each image presentation is static and independent of all those which precede and follow it; the separate presentations were only later combined into an animation. The sequence repeats after the image has made the full circle. Check out the tilt aftereffect demo and direction map demo separately to see the effects of adaptation and motion. A faster-loading version of this page with smaller pictures is also available.

These images illustrate a hypothesis about how the cortex represents and processes orientation. The colorful box (g) shows an orientation map measured for an RF-LISSOM network after self-organization. The orientation map represents a flattened 5mm x 5mm area of the human (or monkey) cortex, specifically the brain area V1. V1 is also called the primary visual cortex, and is a quarter-sized area at the back of the head that is the first cortical stage for most visual processing.

Each neuron in the orientation map is represented by the color of the orientation it prefers, using the color code in (f). The orientation map consists of clearly defined regions preferring each of the possible orientations at a given position on the retina, as seen in the primary visual cortex of experimental animals (e.g. Blasdel 1992). All orientations are approximately equally represented here (because the network was trained on a uniformly random distribution of angles), so the overall histogram of orientation preferences (h) is nearly flat.

The animated panels on the left show the response of this simulated cortex to a single line at the center of the retina, using the same color code for each neuron. The rotating grey line (a) shows the input presented to the retina. The fuzzy colored areas (b) show which neurons responded at the instant the activation reached the cortex, as a result of the afferent connections (from the eyes) only. The histogram adjacent to it (c) shows the total amount of activation in the initial response for orientation-selective neurons of each angle. The sharply defined areas (d) show the neurons still responding after lateral interactions have taken place, after the image has been stable on the retina long enough for the intracortical activity to settle. Finally, the second histogram (e) shows the total amount of activation in the settled response, for each angle of orientation-selective neuron.

The initial response is broad and diffuse, both spatially and in the orientation domain shown in histogram (c). However, there is still a clear orientation preference evident in the response. For instance, when the line is tilted to the left, red, yellow, and green areas are activated, and the peak of the histogram is in this region. When the line is tilted to the right, the blue and purple areas are activated, and the peak of the histogram is in the blue and purple region.

After settling due to the lateral interactions in the model, the response consists of well-defined areas of activation representing neurons which prefer orientations quite close to that of the input pattern. Now only orientations near that of the input pattern show up at all in the histogram (e).

This process encodes orientation as a pattern of activity across orientation selective neurons. The encoding is a sparse distributed code: it is distributed across many neurons, yet it is sparse: only a relatively few number overall are active. Among other advantages, this type of encoding represents a balance between representing the maximum amount of information in the smallest space, and being able to tolerate failures in many of the constituent elements.