Modeling Neural Oscillations Using VLSI-Based Neuromimes
As a prelude to a VLSI implementation of a live locomotory control network, the phenomena of reciprocal inhibition and recurrent cyclic inhibition were recreated and subjected to parametric tests of their oscillatory range and stability. Networks were constructed from comprehensive VLSI-based artificial neurons, or neuromimes, which are efficient and convenient, yet configurable and comprehensive, allowing a variety of cellular transient and steady-state characteristics to be precisely and continuously varied. In initial tests, both oscillator types were found to operate over a broad range of cellular and network frequencies. In those tests, it was noted that oscillation by reciprocal inhibition between two neurons requires that each possess some measure of synaptic dynamics, while neurons in the cyclically inhibited networks did not. This suggests that the two oscillators utilized different temporal mechanisms. In circuit tests, individual cells self-oscillated over a wide range of frequencies in response to cell threshold, refractory period, and postsynaptic inhibition. In subsequent network tests, the cyclic networks were found to be sensitive to cellular threshold, yet insensitive to refractory period duration. Conversely, the reciprocal networks were found to be sensitive to refractory period duration, yet immune to cellular threshold. This complementary relationship suggests an advantage for biological oscillatory networks that incorporate both types.