Memristor Bridge Research Articles

Memristor bridge synapse for better artificial neuron perceptron

In artificial neural networks, the fourth passive element memristor can be utilized as an electronic synapse that serves as the interface between neurons. The artificial neuron composed of the memristor bridge synapse not only has the characteristics of low power consumption and high integration but also has a more simplified circuit and weight change conditions. Particularly, it has the ability of bionic intelligent information processing. This paper established two novel synaptic structures on the basis of memristor bridges (type 1 and type 2) and then synthetically analyzed how to realize the artificial neuron perceptron. Herein, the artificial synapses (type 1 and type 2) have the following characteristics: continuous changes in synaptic weights (positive, negative, and zero) and memory properties. Among them, the type 2 memristor bridge has the advantage of a wider range of weight updates for the synaptic circuit, which can realize the function of the artificial neuron perceptron with less error. This work lays the foundation for the future exploitation of artificial intelligence.

Memristor Bridge Synapse Application for Integrate and Fire and Hodgkin-Huxley Neuron Cell

Memristor Bridge Synapse Application for Integrate and Fire and Hodgkin-Huxley Neuron Cell

Forgetting memristors and memristor bridge synapses with long- and short-term memories

In this paper, an ideal current source forgetting memristor model is proposed. Based on the model, three kinds of synapses with long- and short-term memory are designed: the series forgetting memristor synapse, the forgetting memristor bridge synapse written independently and the forgetting memristor bridge synapse written in batches. Combined with the forgetting property, the long- and short-term weight of the forgetting synapse can be controlled by the long- and short-term resistance of memristors. Compared the three forgetting synapses, the series forgetting memristor synapse has the lowest requirement for memristors, the forgetting memristor bridge synapse written independently is the most flexible, and the forgetting memristor bridge synapse written in batches is the most convenient. Compared with traditional synapses, forgetting synapses with long- and short-term memory have multi-weight storage. When forgetting synapses are applied to associative memory, it can be found more patterns are stored in the neural network and different patterns are recalled at different time due to the forgetting effect.

A reconfigurable bidirectional associative memory network with memristor bridge

Bidirectional associative memories (BAMs) have been extensively applied in auto and heteroassociative learning. However, the research on the hardware implementation of BAM neural networks is relatively few. Memristor, a nanoscale synaptic-like element, provides a new perspective for the circuit implementation of neural networks. In this work, we improve a threshold memristor bridge circuit (T-MBC) and design a novel memristive bidirectional associative memory (MBAM) network circuit on this basis. T-MBC is capable to achieve positive, negative, and zero synaptic weights without any switches, inverters, transistors, and current–voltage conversion processes. The validity of T-MBC is illustrated by PSPICE. MBAM has a simpler structure and gains excellent effects in flexibility, and reconfigurability. Besides, it is shown that MBAM can be trained to recall both disturbed binary images and incomplete/noisy grey-scale images robustly. Meanwhile, a more complex application of emotion classification based on MBAM is also accomplished. Experimental results verify the effectiveness of the MBAM neural network and demonstrate that the MBAM neural network has high recognition accuracy and strong noise immunity.

Low power convolutional architectures: Three operator switching systems based on forgetting memristor bridge

With the development of technology and society, artificial intelligence has entered the era of deep learning neural networks. Convolution operation and image processing are fundamental technologies for deep neural networks and artificial intelligence, so reducing the energy consumption of convolution operation and increasing the speed of image processing will promote the development of a sustainable intelligent society. Traditional image processing technologies are based on the Von Neumann architecture, which is slow and not convenient for the hardware implementation of neural networks. Therefore, this paper breaks through the Von Neumann architecture, and uses the forgetting memristor bridge to realize the parallel image processing on the neumorphic chips. We use the forgetting characteristics of the memristor bridge to switch operators, and design single-operator switching, double-operator switching and K-operator switching three system architectures. The image operator switching systems designed in this paper not only have the similar processing effect of the traditional way, but also have the advantages of easy control, fast running speed (the processing speed is reduced from ms to μs), and low power consumption (the power consumption is reduced by nearly half), which will bring immeasurable benefits to the sustainable intelligent society.

On Learning With Nonlinear Memristor-Based Neural Network and its Replication

Nonlinear memristor-based neural network and a circuit-based learning system is addressed in this work. The weights of the neural network are based on the memristor bridge synapse and the learning architecture is designed in analog–digital mixed circuits by adopting a simple learning algorithm called random weight change algorithm. Though the memristor bridge can be efficiently used as a synapse, it still suffers from nonlinearity in weight programming at its extremes due to the boundary effect of memristors, which is a common phenomenon in most of the nano-devices. In this study, a novel architecture of a modified memristor bridge synapse is proposed that avoids the boundary effect issue by shifting the programming origin to the middle of the linear region. To demonstrate the effectiveness of the proposed method, multilayer neural network along with the end-to-end learning architecture is designed in circuit with nonlinear memristors and tested for several classical learning problems. Simulation results showing successful learning and correct behavior of replicated neural network are presented.

Memristor bridge-based low pass filter for image processing

This paper highlights the memristor bridge-based lowpass filter (LPF) and improved image processing algorithms along with a novel adaptive Gaussian filter for denoising image and a new Gaussian pyramid for scale invariant feature transform (SIFT). First, a novel kind of LPF based on the memristor bridge is designed, whose cut-off frequency and other traits are demonstrated to change with different time and memristance. In light of the changeable parameter of the memristor bridge-based LPF, a new adaptive Gaussian filter and an improved SIFT algorithm are presented. Finally, experiment results show that the peak signalto-noise ratio (PSNR) of our denoising is bettered more than 2.77 dB compared to the corresponding of the traditional Gaussian filter, and our improved SIFT performances including the number of matched feature points and the percent of correct matches are higher than the traditional SIFT, which verifies feasibility and effectiveness of our algorithm.

Rectifying Memristor Bridge Circuit Realized with Human Skin.

It has been demonstrated before that human skin can be modeled as a memristor (memory resistor). Here we realize a memristor bridge by applying two voltages of opposite signs at two different skin sites. By this setup it is possible to use human skin as a frequency doubler and half-wave rectifier which is an application of the non-linear electrical properties of human skin. The corresponding electrical measurements are non-linear since these are affected by the applied stimulus itself.

Excitatory and inhibitory actions of a memristor bridge synapse

Excitatory and inhibitory actions of a memristor bridge synapse

Modular neuron comprises of memristor-based synapse

Design of analog modular neuron based on memristor is proposed here. Since neural networks are built by repetition of basic blocks that are called neurons, using modular neurons is essential for the neural network hardware. In this work modularity of the neuron is achieved through distributed neurons structure. Some major challenges in implementation of synaptic operation are weight programmability, weight multiplication by input signal and nonvolatile weight storage. Introduction of memristor bridge synapse addresses all of these challenges. The proposed neuron is a modular neuron based on distributed neuron structure which it uses the benefits of the memristor bridge synapse for synaptic operations. In order to test appropriate operation of the proposed neuron, it is used in a real-world application of neural network. Off-chip method is used to train the neural network. The results show 86.7 % correct classification and about 0.0695 mean square error for 4-5-3 neural network based on proposed modular neuron.

신경회로망용 멤리스터 브릿지 회로에서 가중치 프로그램의 시간에 대한 선형화 효과

Memristor is a new kind of memory device whose resistance varies depending upon applied charge and whose previous resistance state is preserved even when its power is off. Ordinary memristor has a nonlinear programming characteristics about time when a constant voltage is applied. For the easiness of programming, it is desirable that resistance is programmed linearly about time. We had proposed previously a memristor bridge configuration with which weight can be programmed nicely in positive, negative or zero. In memristor bridge circuit, two memristors are connected in series with different polarity. Memristors are complementary each other and it follows that the memristance variation is linear with respect to time. In this paper, the linearization effect of weight programming of memristor bridge synapse is investigated and verified about both memristor from HP and a nonlinear memristor with a window function. Memristor bridge circuit would be helpful to conduct synaptic weight programming.

A Circuit-Based Learning Architecture for Multilayer Neural Networks With Memristor Bridge Synapses

Memristor-based circuit architecture for multilayer neural networks is proposed. It is a first of its kind demonstrating successful circuit-based learning for multilayer neural network built with memristors. Though back-propagation algorithm is a powerful learning scheme for multilayer neural networks, its hardware implementation is very difficult due to complexities of the neural synapses and the operations involved in the learning algorithm. In this paper, the circuit of a multilayer neural network is designed with memristor bridge synapses and the learning is realized with a simple learning algorithm called Random Weight Change (RWC). Though RWC algorithm requires more iterations than back-propagation algorithm, we show that a circuit-based learning using RWC is two orders faster than its software counterpart. The method to build a multilayer neural network using memristor bridge synapses and a circuit-based learning architecture of RWC algorithm is proposed. Comparison between software-based and memristor circuit-based learning are presented via simulations.

Frequency dependent rectifier memristor bridge used as a programmable synaptic membrane voltage generator

Abstract Reasoned by its dynamical behavior, the memristor enables a lot of new applications in analog circuit design. Since some realizations have been shown (e.g. 2007, Hewlett Packard), the development of applications with memristors becomes more and more interesting. Besides applications in neural networks and storage devices, analog memristive circuits also promise further applications. Therefore, this article proposes a frequency dependent rectifier memristor bridge for different purposes, for example, using as a programmable synaptic membrane voltage generator for Spike-Time-Dependent-Plasticity and describes the circuit theory. In this context it is shown that the Picard Iteration is one possibility to analytically solve the system of nonlinear state equations of memristor circuits. An intuitive picture of how a memristor works in a network in general is given as well and in this context some research on the dynamical behavior of a HP memristor should be done. After all it is suggested to use the memristor bridge as a neuron.

Memristor Bridge Synapse-Based Neural Network and Its Learning

Analog hardware architecture of a memristor bridge synapse-based multilayer neural network and its learning scheme is proposed. The use of memristor bridge synapse in the proposed architecture solves one of the major problems, regarding nonvolatile weight storage in analog neural network implementations. To compensate for the spatial nonuniformity and nonideal response of the memristor bridge synapse, a modified chip-in-the-loop learning scheme suitable for the proposed neural network architecture is also proposed. In the proposed method, the initial learning is conducted in software, and the behavior of the software-trained network is learned by the hardware network by learning each of the single-layered neurons of the network independently. The forward calculation of the single-layered neuron learning is implemented on circuit hardware, and followed by a weight updating phase assisted by a host computer. Unlike conventional chip-in-the-loop learning, the need for the readout of synaptic weights for calculating weight updates in each epoch is eliminated by virtue of the memristor bridge synapse and the proposed learning scheme. The hardware architecture along with the successful implementation of proposed learning on a three-bit parity network, and on a car detection network is also presented.

Memristor Bridge Synapses

In this paper, we propose a memristor bridge circuit consisting of four identical memristors that is able to perform zero, negative, and positive synaptic weightings. Together with three additional transistors, the memristor bridge weighting circuit is able to perform synaptic operation for neural cells. It is compact as both weighting and weight programming are performed in a memristor bridge synapse. It is power efficient, since the operation is based on pulsed input signals. Its input terminals are utilized commonly for applying both weight programming and weight processing signals via time sharing. In this paper, features of the memristor bridge synapses are investigated using the TiO memristor model via simulations.

A voltage mode memristor bridge synaptic circuit with memristor emulators.

A memristor bridge neural circuit which is able to perform signed synaptic weighting was proposed in our previous study, where the synaptic operation was verified via software simulation of the mathematical model of the HP memristor. This study is an extension of the previous work advancing toward the circuit implementation where the architecture of the memristor bridge synapse is built with memristor emulator circuits. In addition, a simple neural network which performs both synaptic weighting and summation is built by combining memristor emulators-based synapses and differential amplifier circuits. The feasibility of the memristor bridge neural circuit is verified via SPICE simulations.