Using the GasNet model in discrete domains

A neural network model-the GasNet-has been reported in the literature, which, in addition to the traditional electric type, point-to-point communication between units, also uses communication through a diffusable chemical modulator. Here we assess the applicability of this model in two different scenarios, the XOR problem and a simulated food gathering task. Both represent simpler and more discrete domains than the one in which GasNet was originally introduced (which had an essentially continuous nature), thus allowing for distinct issues to be addressed; also, both are well-known benchmark problems from the literature. The experiments were intended to better understand the model from analogies with traditional architectures as well as to extend the original problem domain, comparing its performance with some of the ones previously presented.