The article published as Early Access!
Systematic parameter selection for a spiking neuron model based on a volatile memristor and a tunnel diode
Purpose. This study aims to generalize an artificial spiking neuron model to accommodate volatile memristors with arbitrary characteristics and to develop a coordinated procedure for selecting the remaining circuit parameters for a given memristor model. A nanocomposite device based on poly-para-xylylene with molybdenum trioxide nanoparticles is considered as an example. Its parameters are held fixed while the characteristics of the diode branch, bias sources, and charge-storage capacitance are varied.
Methods. The study combines analysis of the current–voltage characteristics of individual circuit elements and the combined current–voltage characteristics of parallel branches, numerical determination of switching times at a fixed voltage, identification of a nonlinear characteristic using normalized residuals, and integration of differential equations. Static conditions for reaching the turn-on region and subsequent recovery are compared with the behavior of the full dynamical model.
Results. A procedure is formulated that relates the characteristics of the switching device to the current threshold for excitation, the admissible input-current range, and the choice of capacitance for the neuron circuit. Three parameter sets with different depths and extents of the negative differential resistance region are examined for the selected memristor. Numerical simulations confirm periodic spike generation under constant excitation and spiking activity under rectangular and sinusoidal inputs. The role of finite switching times in the differences between static and dynamic boundaries is demonstrated. A region of chaotic dynamics under periodic forcing is examined.
Conclusion. The proposed procedure enables coordinated selection of neuron circuit elements without changing the parameters of the selected switching device. Applicability to other devices depends on the existence of compatible current and time ranges: static analysis narrows the search, whereas transient simulations establish the viability of a particular parameter set. The procedure applies to analytical models and experimentally specified characteristics, supports the assessment of parameter variability, and identifies the need to modify the topology when the elements are incompatible.
-
- 78 reads