This article provides a review of current development and challenges in brain-inspired computing with memristors. We review the mechanisms of various memristive devices that can mimic synaptic and neuronal functionalities and survey the progress of memristive spiking and artificial neural networks. Different architectures are compared, including spiking neural networks, fully connected artificial neural networks, convolutional neural networks, and Hopfield recurrent neural networks. Challenges and strategies for nanoelectronic brain-inspired computing systems, including device variations, training, and testing algorithms, are also discussed.
10 August 2017 in IEEE Transactions on Circuits and Systems I: Regular Papers
Memristors are promising components for applications in nonvolatile memory, logic circuits, and neuromorphic computing. In this paper, a novel circuit for memristor-based multilayer neural networks is presented, which can use a single memristor array to realize both the plus and minus weight of the neural synapses. In addition, memristor-based switches are utilized during the learning process to update the weight of the memristor-based synapses. Moreover, an adaptive back propagation algorithm suitable for the proposed memristor-based multilayer neural network is applied to train the neural networks and perform the XOR function and character recognition. Another highlight of this paper is that the robustness of the proposed memristor-based multilayer neural network exhibits higher recognition rates and fewer cycles as compared with other multilayer neural networks.
1 September 2016 in IEEE Transactions on Circuits and Systems II: Express Briefs
As a promising alternative for next-generation memory, memristors provide several useful features such as high density, nonvolatility, low power, and good scalability as compared with conventional CMOS-based memories. In this brief, a voltage-controlled threshold memristive model is proposed, which is based on experimental data of memristive devices. Moreover, the model is more suitable for the design of memristor-based synaptic circuits as compared with other memristive models. The effects of memristance variations are considered in the proposed model to evaluate the behavior of memristive synapses within memristor-based neural networks.
1 March 2017 in IEEE Transactions on Electron Devices
The memristor, a promising candidate for synaptic interconnections in artificial neural network, has gained significant attention for application to neuromorphic systems. One common method is using two memristors as one synapse to represent the positive and negative weights. In this paper, the synaptic behavior of a Ag/AgInSbTe/Ta (AIST)-based memristor is experimentally demonstrated. In addition, a neural architecture using one AIST memristor as a synapse is proposed, where both the plus and minus weights of the neural synapses are realized in a single memristive array. Moreover, the memristor-based neural network is extended to a multilayer architecture, and modified memristor-based backpropagation learning rules are implemented on-chip to achieve pattern recognition. The effects of device variations and input noise on the performance of a memristor-based multilayer neural network (MNN) are also described. The proposed MNN is capable of pattern recognition with high success rates and exhibits several advantages, such as good accuracy, high robustness, and noise immunity.
6 April 2015 in IEEE Transactions on Circuits and Systems I: Regular Papers
Recently, it has been demonstrated that memristors can be utilized as logic gates, control switches as well as memory elements. In this paper, we analyze the different AND, OR, and NOT logic gates which are based on memristors. In addition, a novel design for a memristor-based switch is presented, which can be used in the peripheral read/write circuits of the memristor-based memory. Moreover, methods of consecutive read with long refresh intervals and fast write for the proposed design are also discussed. Another highlight of this work is the analysis of the proposed memristor-based crossbar architecture which has a series of excellent features, such as good-compatibility, high-density, non-volatility, low-power, and good-scalability. Simulation results also show that the proposed memory array has superior performances compared to other memristor-based arrays proposed in the existing technical literature.
Existing deep neural networks (DNNs) are computationally expensive and memory intensive, which hinder their further deployment in novel nanoscale devices and applications with lower memory resources or strict latency requirements. In this paper, a novel approach to accelerate on-chip learning systems using memristive quantized neural networks (M-QNNs) is presented. A real problem of multilevel memristive synaptic weights due to device-to-device (D2D) and cycle-to-cycle (C2C) variations is considered. Different levels of Gaussian noise are added to the memristive model during each adjustment. Another method of using memristors with binary states to build M-QNNs is presented, which suffers from fewer D2D and C2C variations compared with using multilevel memristors. Furthermore, methods of solving the sneak path issues in the memristive crossbar arrays are proposed. The M-QNN approach is evaluated on two image classification datasets, that is, ten-digit number and handwritten images of mixed National Institute of Standards and Technology (MNIST). In addition, input images with different levels of zero-mean Gaussian noise are tested to verify the robustness of the proposed method. Another highlight of the proposed method is that it can significantly reduce computational time and memory during the process of image recognition.