Razvan V. Florian, Dumitru Dumitrescu
We study the possibility of developing far-from-equilibrium computational systems, in analogy with corresponding self-organizing physical systems. We review the computational models developed in analogy with physical systems converging towards equilibrium and the research related to the self-organization of far-from-equilibrium physical systems. Based on this review, we define several nonequilibrium computational models. We consider a classification of the state of a nonequilibrium informational system by a linear readout as defining an observer-dependent nonequilibrium steady state. We thus reinterpret the liquid state machine framework as a specific nonequilibrium computational model, which permits its study under new perspectives. We also define two new types of nonequilibrium computational models that may lead to useful applications, similar to the liquid state machine.