An ultra-compact leaky-integrate-and-fire model for building spiking neural networks

Abstract

We introduce an ultra-compact electronic circuit that realizes the leaky-integrate-and-fire model of artificial neurons. Our circuit has only three active devices, two transistors and a silicon controlled rectifier (SCR). We demonstrate the implementation of biologically realistic features, such as spike-frequency adaptation, a refractory period and voltage modulation of spiking rate. All characteristic times can be controlled by the resistive parameters of the circuit. We built the circuit with out-of-the-shelf components and demonstrate that our ultra-compact neuron is a modular block that can be associated to build multi-layer deep neural networks. We also argue that our circuit has low power requirements, as it is normally off except during spike generation. Finally, we discuss the ultimate ultra-compact limit, which may be achieved by further replacing the SCR circuit with Mott materials.

A spiking and adapting tactile sensor for neuromorphic applications

“…The IVcharacteristic of this pair is strongly nonlinear, with a negative differential resistance. Similar characteristics can be obtained using for example thyristors, programmable unijunction transistors or highly correlated electron materials such as VO 2 [60][61][62] . As the applied voltage exceeds the threshold voltage of these transistors, the current increases.…”

Biologically Relevant Dynamical Behaviors Realized in an Ultra-Compact Neuron Model

“…In the present work we shall describe how the functionality of the basic UCN block can be extended to realize a variety of biologically relevant spiking patterns, without a sacrifice of circuit simplicity. The paper is organized as follows: In section Materials and Methods we shall describe our recently introduced UCN circuit (Rozenberg et al, 2019). We shall demonstrate how the basic behavior of the UCN can be very precisely captured by means of numerical simulations obtained with LTspice (LTspice R , 2020 that we validate against actual circuit measured data.…”

“…Namely the threshold switching of the SCR conductance emulates the firing of biological neurons. As we discussed in Rozenberg et al (2019), this features permits a drastic reduction to the number of components to implement a basic leaky-integrateand-fire (LIF) SiN. So in this regard it may be considered as belonging to the class of Compact SiN circuits (Indiveri et al, 2011).…”

“…In a recent paper (Rozenberg et al, 2019) we introduced the ultracompact spiking neuron model that only requires two transistors and a thyristor (SCR). We follow the terminology of Indiveri et al (2011), where a compact model refers to a simple electronic circuit with few components.…”

“…Interestingly, the same dilemma is encountered for the implementation of neurons in neuromorphic circuits -i.e., circuits that perform computations based on the architecture of the brain. Recently we proposed an ultra-compact neuron (UCN) circuit that realizes the leaky integrate and fire model (Rozenberg et al, 2019). That circuit complies with the first requirement, as it was simply based on only three active devices, two transistors and a thyristor, or SCR, plus a capacitor and a few resistors.…”