A neuromorphic systems approach to in-memory computing with non-ideal memristive devices: from mitigation to exploitation

Abstract

Memristive devices represent a promising technology for building neuromorphic electronic systems. In addition to their compactness and non-volatility features, they are characterized by computationally relevant physical properties, such as state-dependence, non-linear conductance changes, and intrinsic variability in both their switching threshold and conductance values, that make them ideal devices for emulating the bio-physics of real synapses. In this paper we present a spiking neural network architecture that supports the use of memristive devices as synaptic elements, and propose mixed-signal analog-digital interfacing circuits which mitigate the effect of variability in their conductance values and exploit their variability in the switching threshold, for implementing stochastic learning. The effect of device variability is mitigated by using pairs of memristive devices configured in a complementary push-pull mechanism and interfaced to a current-mode normalizer circuit. The stochastic learning mechanism is obtained by mapping the desired change in synaptic weight into a corresponding switching probability that is derived from the intrinsic stochastic behavior of memristive devices. We demonstrate the features of the CMOS circuits and apply the architecture proposed to a standard neural network hand-written digit classification benchmark based on the MNIST data-set. We evaluate the performance of the approach proposed on this benchmark using behavioral-level spiking neural network simulation, showing both the effect of the reduction in conductance variability produced by the current-mode normalizer circuit, and the increase in performance as a function of the number of memristive devices used in each synapse.Neuromorphic computing systems comprise synapse and neuron circuits arranged in a massively parallel manner to support the emulation of large-scale spiking neural networks 1-9 . In many of these systems, and in particular in neuromorphic processing devices designed to overcome the von-Neumann bottleneck problem 7,8,10-14 , the bulk of the silicon real-estate is taken up by synaptic circuits that integrate in the same area both memory and computational primitives. To save area and maximize density in such devices, one possible approach is to implement very basic synapse circuits arranged in dense cross-bar arrays [15][16][17][18][19] . However, such approach is likely to relegate the role of the synapse to a basic multiplier 14,20 . In biology, synapses are extremely sophisticated structures that exhibit complex and powerful computational properties, including temporal dynamics, state-dependence, and stochastic learning behavior. The challenge is to design neuromorphic circuits that emulate these computational properties, and are also compact and low power. Memristive devices have recently emerged as nano-scale devices which provide a promising technology for addressing these problems 31,46 . These devices offer a compact and efficient solution to model synaptic weights since they are non-volatile, have ...

Memristors for the Curious Outsiders

“…The impact of the dynamical model of the memristor has been studied in the implementation of STDP, and the learning rules can be adapted to the different behaviors. As in most application of memristors as non-volatile storage of (synaptic) weights, it suffers from the intrinsic variability of the units, which more general neuromorphic circuits are able to exploit [64].…”

Memristive Spiking Neural Networks Trained with Unsupervised STDP

“…[28] is adopted. Though our proposed architecture and learning method are demonstrated by simulation, the used models and parameters are set according to experimental results [1,2,7,11]. Parameters for simulation are shown in Table 1.…”

“…These systems use memristors as synapses to store synaptic weights and implement multiplication and addition by physical processes, which are energy-efficient. For instance, Payvand et al proposed a neuromorphic system with non-ideal memristive devices and demonstrated the system using behavioral simulation [7]. There are mainly two kinds of memristive neuromorphic computing systems: voltage-based systems and spike-based systems [1,2,[8][9][10][11].…”

Complementary Metal‐Oxide Semiconductor and Memristive Hardware for Neuromorphic Computing

“…SNN and artificial neural network systems that utilize memristors as their learning components or simply as a programmable element have been researched extensively . Here, we mainly focus on neuromorphic systems that utilize SiO x memristive devices for learning, where memristors are used as weight elements.…”