Biologically inspired evolutionary temporal neural circuits

The success of classical artificial neural networks as well as the wealth of new data about neuronal functions encourages researchers to explore new possibilities for the creation of biologically inspired artificial neural circuits. Temporal sequence is an important factor which is underrepresented in classical artificial neural networks. We developed a computational model of a neuronal circuit that incorporates more biological resemblance in a form of a fully interconnected temporal network. The proposed function of the circuit is crucial in selecting the parameters which control each neuron and their final arrangement. However, the mathematics behind the dynamics of nonlinear, time sensitive, interconnected neural circuits and their learning processes as well as the choice of temporal neural code is daunting. To solve this problem, we use evolutionary optimization methods to find the appropriate configuration of simulated neural circuits. The results of a preliminary simulated neural circuit demonstrate that pattern recognition tasks can be performed with robustness to noisy or distorted patterns.