The phasic “brightness” neuron extracts the change in brightness from the visual input. The working memory neurons in the second layer are recurrent and thus maintain exponentially decaying memory traces. The “brightness working memory” neuron receives strong input from the “brightness” neuron, and signals recent changes in brightness. The “counting working memory” neuron receives weak input from the “brightness” neuron, and so accumulates information about the changes in brightness over a longer period. Finally, the “evaluation” neuron subtracts the “counting working memory” from the “brightness working memory.” Its response is inversely proportional to the number of brightness changes, and, with the right visual input, it provides an online evaluation of the numerosity of the stimulus. From: Vasas & Chittka (2018) Insect-Inspired Sequential Inspection Strategy Enables an Artificial Network of Four Neurons to Estimate Numerosity
iScience 11:85-92.