4

Many years ago I saw an experiment which involved a lab primate whose brain was monitored as it watched a simple rectangular object as it moved. And sure enough, there were regions of cells in its brain that seemed, IIRC, to sequentially fire in some manner that corresponded to the rectangle. I do not recall if it was merely speed or if there was actually a rectangular region that "moved" (that is, different cells would fire in adjacent regions).

So my question is whether or not this sort of thing might be observed in some kinds of artificial minds. If, for example, you asked it to display an object as it rotated would there be some set of artificial neurons that would be allocated to represent the object and rotated in the same sense as described above, that is, artificial neurons would gradually "fire" in adjacent regions during the rotation?

nbro
  • 42,615
  • 12
  • 119
  • 217
Jeff
  • 143
  • 4

1 Answers1

4

Since ANNs are inspired from various types of biological neural coding schemes, it won't be too surprised to see the similar activation saliency map between them. One reference is Li et al's (2022) Artificial Visual System for Orientation Detection Based on Hubel–Wiesel Model.

The Hubel–Wiesel (HW) model is a classical neurobiological model for explaining the orientation selectivity of cortical cells... To realize a straightforward and efficient quantitive method and validate the HW model’s reasonability and practicality, we use McCulloch-Pitts (MP) neuron model to simulate simple cells and complex cells and implement an artificial visual system (AVS) for two-dimensional object orientation detection... Computer simulations show that the mechanism-based AVS can make accurate orientation discrimination and shows striking biological similarities with the natural visual system, which indirectly proves the rationality of the Hubel–Wiesel model.

Another example could be Hinton's CapsNet.

A capsule is a set of neurons that individually activate for various properties of a type of object, such as position, size and hue. Formally, a capsule is a set of neurons that collectively produce an activity vector with one element for each neuron to hold that neuron's instantiation value (e.g., hue)... For low level capsules, location information is "place-coded" according to which capsule is active. Higher up, more and more of the positional information is rate-coded in the capsule's output vector... Capsnets use neural activities that vary with viewpoint.

cinch
  • 11,000
  • 3
  • 8
  • 17