Memristor Crossbar-Based Neuromorphic Computing System: A Case Study

Abstract

By mimicking the highly parallel biological systems, neuromorphic hardware provides the capability of information processing within a compact and energy-efficient platform. However, traditional Von Neumann architecture and the limited signal connections have severely constrained the scalability and performance of such hardware implementations. Recently, many research efforts have been investigated in utilizing the latest discovered memristors in neuromorphic systems due to the similarity of memristors to biological synapses. In this paper, we explore the potential of a memristor crossbar array that functions as an autoassociative memory and apply it to brain-state-in-a-box (BSB) neural networks. Especially, the recall and training functions of a multianswer character recognition process based on the BSB model are studied. The robustness of the BSB circuit is analyzed and evaluated based on extensive Monte Carlo simulations, considering input defects, process variations, and electrical fluctuations. The results show that the hardware-based training scheme proposed in the paper can alleviate and even cancel out the majority of the noise issue.

Lag quasi-synchronization for memristive neural networks with switching jumps mismatch

“…Since the pioneering work by Chua [1], who originally proposed the memristor concept, the research of memristor has become an interesting issue [2][3][4][5]. By replacing the resistors in the artificial neural networks [6][7][8] with memristors, MNNs can be constructed.…”

The applications of memristor devices in next-generation cortical processor designs

“…Fig. 5 shows the diagram of on-chip training scheme for the proposed memristor crossbar design based on the extend Delta rule [12], [13]. It is used to find the weights that minimize the squared error between a target output pattern (e.g., character "a") and the input prototype patterns (e.g., a group of given images of "a" in different sizes/fonts/styles).…”

Practical gradient-descent for memristive crossbars

“…It allows fast feed-forward fully parallel on-line hardware based learning, without requiring accurate models of the memristor behaviour and precise control of the programming pulses. The effect of device parameters, training parameters, and device variability on the learning performance of crossbar arrays trained using the USD algorithm has been studied via simulations.There is a significant interest in using memristive devices for computation, in particular in the context of neuromorpic systems [1] and artificial neural networks [2][3][4][5][6][7]. Memristors are typically fabricated in the form of highly-dense crossbar arrays, which naturally lend themselves to the vector-matrix multiplications that are at the core of the neural network algorithms.…”

“…There is a significant interest in using memristive devices for computation, in particular in the context of neuromorpic systems [1] and artificial neural networks [2][3][4][5][6][7]. Memristors are typically fabricated in the form of highly-dense crossbar arrays, which naturally lend themselves to the vector-matrix multiplications that are at the core of the neural network algorithms.…”

“…We are also interested in in situ training, where the crossbar array itself is used to implement the learning algorithm, as opposed to ex situ training, where the weights are determined outside the array, and then uploaded to the memristors. In situ training has many advantages, the most important of which is in allowing the array to any defects and hardware variability issues [3][4][5].…”