FRQI Pairs method for image classification using Quantum Recurrent Neural Network

Classical Convolutional Neural Networks, or ConvNets, have been the benchmarks for most object classification and face recognition tasks despite suffering from major limitations such as the inability to capture spatial co-locality between data points and favoring invariance over equivariance. Hence, a hierarchical routing layered architecture called Capsule Networks was proposed to overcome shortcomings of ConvNets. Capsules replace average or max pooling techniques of ConvNets with dynamic routing abilities between lower level and higher level neural units which better capture hierarchical relationships within the data and introduced reconstruction regularization mechanisms which deals with equivariance properties. By overcoming existing limitations, Capsules have proven themselves to be potential benchmarks in object segmentation, detection and reconstruction. Capsules have achieved state of the art results on the fundamental MNIST (Modified National Institute of Standards and Technology) handwritten digit dataset by reducing the ConvNets test error benchmark of 0.39% to 0.25%. In order to further augment this distinction, we experimented with five activation units such as sigmoid, e-Swish, Swish, variants of Rectified Linear Units (ReLU) like Parametric ReLU (PReLU), leaky ReLU (lReLU) and Scaled Exponential Linear Units (SELU), on two fundamental datasets - MNIST and cifar10. Based on these experimental results, we establish that e-Swish, and ReLU variants better optimize the Capsule architecture as compared to the currently used ReLU activation function.

Owing to the controllable growth and large-area synthesis for high-density integration, interest in employing atomically thin two-dimensional (2D) transition-metal dichalcogenides (TMDCs) for synaptic transistors is increasing. In particular, substitutional doping of 2D materials allows flexible modulation of material physical properties, facilitating precise control in defect engineering for eventual synaptic plasticity. In this study, to increase the switch ratio of synaptic transistors, we selectively performed experiments on WS2 and introduced niobium (Nb) atoms to serve as the channel material. The Nb atoms were substitutionally doped at the W sites, forming a uniform distribution across the entire flakes. The synaptic transistor devices exhibited an improved switch ratio of 103, 100 times larger than that of devices prepared with undoped WS2. The Nb atoms in WS2 play crucial roles in trapping and detrapping electrons. The modulation of channel conductivity achieved through the gate effectively simulates synaptic potentiation, inhibition, and repetitive learning processes. The Nb-WS2 synaptic transistor achieves 92.30% recognition accuracy on the Modified National Institute of Standards and Technology (MNIST) handwritten digit dataset after 125 training iterations. This study’s contribution extends to a pragmatic and accessible atomic doping methodology, elucidating the strategies underlying doping techniques for channel materials in synaptic transistors.

We assess the benefit of including an image inpainting filter before passing damaged images into a classification neural network. We employ an appropriately modified Cahn–Hilliard equation as an image inpainting filter which is solved numerically with a finite-volume scheme exhibiting reduced computational cost and the properties of energy stability and boundedness. The benchmark dataset employed is Modified National Institute of Standards and Technology (MNIST) dataset, which consists of binary images of handwritten digits and is a standard dataset to validate image-processing methodologies. We train a neural network based on dense layers with MNIST, and subsequently we contaminate the test set with damages of different types and intensities. We then compare the prediction accuracy of the neural network with and without applying the Cahn–Hilliard filter to the damaged images test. Our results quantify the significant improvement of damaged-image prediction by applying the Cahn–Hilliard filter, which for specific damages can increase up to 50% and is advantageous for low to moderate damage.

Neuromorphic computing [1], inspired by the energy-efficient and parallelized architecture of the human brain, represents a paradigm shift in computing. This approach not only drives innovative hardware designs but also inspires novel network architectures like spiking neural networks (SNNs), which process information as discrete spike events. However, scalability in neuromorphic systems is constrained by the large hardware footprint. Realizing multimodal functionality using the same set of hardware primitives can offer a promising solution to these limitations.In this work, we present a dual encoder-neuron circuit based on a nanoporous graphene (NPG) memristor device [2]. The graphene leaky integrate-and-fire (GLIF) circuit has a versatile multimodal design that combines the roles of a LIF neuron and a spike encoder within the same architecture. This design mimics the dual functionality of biological neurons, which encode and process signals locally. The spike encoding capability of the circuit is further revealed using single and double layer SNN networks for pattern recognition tasks using the Modified National Institute of Standards and Technology (MNIST) dataset.Inset of Figure 1(a) shows the schematic of the fabricated NPG device with a channel length of 100 µm. The device has a lateral structure with NPG channel and gold (Au) electrodes. Figure 1(a) shows the current-voltage (I-V) characteristics of the fabricated NPG device which shows a threshold switching behavior with a wide hysteresis window and a threshold voltage (Vth) of 4.9 V. A behavioral SPICE model is developed based on a voltage-controlled switch model. It is seen from Figure 1(a) that the I-V curves simulated using the SPICE model matches the experimentally observed device switching characteristics. This reveals that the simplified SPICE model can be a good substitute for complex Verilog based models. Based on the SPICE model, we designed a simple neuron circuit with a resistor (Rs, 12 kΩ) in series with a parallel combination of a capacitor (Cm, 10 nF) and the NPG device (SPICE model). Figure 1(b) shows the variation of the membrane potential (Vm) and the output spike current in response to the input pulse voltage (Vin, 6 V). Vm shows a clear leaky integration behavior and full reset to zero voltage after the spike response. These are essential attributes of a bioplausible LIF neuron circuit. In order to develop the multimodal GLIF circuit, the output spike frequency of the LIF neuron circuit is plotted as a function of the Rs values. This mapping allows pixel values from MNIST input images to be directly translated into normalized conductance values (G/G0). Hence, each pixel is represented as a normalized conductance value, establishing a clear correlation between pixel intensity and spike frequency. This core idea is further used to develop the GLIF encoder based on the same neuron circuit as before. Based on the dual encoder-neuron circuit, we designed a single layer SNN for recognising the handwritten digits from the MNIST dataset. Figure 1(c) shows the spike-encoded images for the MNIST digits “2” and “5”. The probability of spiking in each timestep is proportional to the spiking frequency. The spike encoded images at each time step reveal the contours of the original input images using sparse representation. These sparse matrices highlight the energy efficiency and computational advantages of the GLIF system. The performance of the dual encoder-neuron system is evaluated by training the single layer SNN using both the GLIF encoder and neuron circuit as integral components. The single layer SNN recorded a high pattern recognition accuracy of 90.77% which is comparable to the purely software-based implementation (92.69%). To explore the scalability of the GLIF system, we implemented a double-layer SNN using the GLIF system, achieving a high recognition accuracy of 97.37%. The GLIF circuit exemplifies how reconfigurable primitive components can drive the development of scalable neuromorphic hardware.

A radial basis function network (RBF) has excellent generalization ability and approximation accuracy when its parameters are set appropriately. However, when relying only on traditional methods, it is difficult to obtain optimal network parameters and construct a stable model as well. In view of this, a novel radial basis neural network (RBF-MLP) is proposed in this article. By connecting two networks to work cooperatively, the RBF’s parameters can be adjusted adaptively by the structure of the multi-layer perceptron (MLP) to realize the effect of the backpropagation updating error. Furthermore, a genetic algorithm is used to optimize the network’s hidden layer to confirm the optimal neurons (basis function) number automatically. In addition, a memristive circuit model is proposed to realize the neural network’s operation based on the characteristics of spin memristors. It is verified that the network can adaptively construct a network model with outstanding robustness and can stably achieve 98.33% accuracy in the processing of the Modified National Institute of Standards and Technology (MNIST) dataset classification task. The experimental results show that the method has considerable application value.

Despite its prevalence in neurosensory systems for pattern recognition, event detection, and learning, the effects of sensory adaptation (SA) are not explored in reservoir computing (RC). Monazomycin‐based biomolecular synapse (MzBS) devices that exhibit volatile memristance and short‐term plasticity with two strength‐dependent modes of response are studied: facilitation and facilitation‐then‐depression (i.e., SA). Their ability to perform RC tasks including digit recognition, nonlinear function learning, and aerodynamic gust classification via combination of model‐based device simulations and physical experiments where SA presence is controlled is studied. Simulations exhibiting moderate SA achieve significantly higher accuracy classifying a custom 5 × 5 binary digit set, with experimental validation achieving maximum testing accuracies of 90%. Classifications of the Modified National Institute of Standards and Technology (MNIST) handwritten digit dataset achieve a maximum testing accuracy of 94.34% in devices with SA. Fitting error of the Mackey–Glass time series is also significantly reduced by SA. Experimentally obtained pressure distributions representing gusts on an airfoil in a wind tunnel are classified by MzBS reservoirs. Reservoirs exhibiting SA achieve 100% accuracy, unlike MzBS reservoirs without SA and comparable static neural networks.

Model optimization is a problem of great concern and challenge for developing an image classification model. In image classification, selecting the appropriate hyperparameters can substantially boost the model’s ability to learn intricate patterns and features from complex image data. Hyperparameter optimization helps to prevent overfitting by finding the right balance between complexity and generalization of a model. The ensemble genetic algorithm and convolutional neural network (EGACNN) are proposed to enhance image classification by fine-tuning hyperparameters. The convolutional neural network (CNN) model is combined with a genetic algorithm GA) using stacking based on the Modified National Institute of Standards and Technology (MNIST) dataset to enhance efficiency and prediction rate on image classification. The GA optimizes the number of layers, kernel size, learning rates, dropout rates, and batch sizes of the CNN model to improve the accuracy and performance of the model. The objective of this research is to improve the CNN-based image classification system by utilizing the advantages of ensemble learning and GA. The highest accuracy is obtained using the proposed EGACNN model which is 99.91% and the ensemble CNN and spiking neural network (CSNN) model shows an accuracy of 99.68%. The ensemble approaches like EGACNN and CSNN tends to be more effective as compared to CNN, RNN, AlexNet, ResNet, and VGG models. The hyperparameter optimization of deep learning classification models reduces human efforts and produces better prediction results. Performance comparison with existing approaches also shows the superior performance of the proposed model.

Spiking neural networks (SNNs) have garnered significant attention due to their potential for low energy consumption, fewer hyperparameters, and reduced network complexity compared to deep neural networks. However, the performance of SNNs is largely constrained by the limitations of spiking neurons, including high energy consumption, low operational speed and weak readout signals. Designing an efficient spiking neuron is thus critical for advancing SNN capabilities. In this work, we present a compact and energy-efficient neuron device with self-reset and leaky-integrate-and-fire (LIF) characteristics, enabled by a single spin-orbit torque magnetic tunnel junction (SOT-MTJ). The LIF functionality is realized within a few nanoseconds by leveraging current induced SOT effect and thermal accumulation. The firing probability can be finely controlled by modulating the SOT current, spike width, and spike interval with only 49 fJ for each spike. Moreover, the device autonomously resets after firing due to the stray field engineered from the reference layer. Using this experimentally validated neuron device, we construct a three-layer SNN to classify the Modified National Institute of Standards and Technology (MNIST) dataset, achieving an accuracy of 96.9%. This system showcases exceptional performance in recognition and classification tasks, underscoring the potential of spintronic devices for energy-efficient SNN applications.

Efficient approximate multipliers utilizing compact and low-power compressors for error-resilient applications

Aiming at the problem that traditional image rendering methods are time-consuming and complex and cannot meet the application scenarios of modern design. In this paper, the depth self-encoder and capsule network in artificial intelligence (AI) technology are used to study the automatic rendering of images. Firstly, Scharr filter is used to reconstruct the image in Stack Capsule Autoencoder (SCAE) model to enhance the accuracy of image target detection and reduce the loss of reconstructed image. Then the loss function of the model is improved. Finally, in the Modified National Institute of Standards and Technology (MNIST) data set, Canadian Institute for Advanced Research-10 (CIFAR-10) data set, the performance of the improved model is tested on the Canadian Institute for Advanced Research-100 (CIFAR-100) data set and the data set made by the author. The test results show that the accuracy of the improved model is higher than that of the traditional K-means, Autoencoder network (AE) unsupervised algorithm and the improved depth clustering unsupervised model. The stacked capsule self-encoder using Scharr filter can make the classification effect of the model more accurate. The improved algorithm in this paper provides some references for improving the efficiency of image rendering in virtual environment.

In this study, the development and characterization of 2D ferroelectric field-effect transistor (2D FeFET) devices are presented, utilizing nanoscale ferroelectric HfZrO2 (HZO) and 2D semiconductors. The fabricated device demonstrated multi-level data storage capabilities. It successfully emulated essential biological characteristics, including excitatory/inhibitory postsynaptic currents (EPSC/IPSC), Pair-Pulse Facilitation (PPF), and Spike-Timing Dependent Plasticity (STDP). Extensive endurance tests ensured robust stability (107 switching cycles, 105 s (extrapolated to 10 years)), excellent linearity, and high Gmax /Gmin ratio (>105 ), all of which are essential for realizing multi-level data states (>7-bit operation). Beyond mimicking synaptic functionalities, the device achieved a pattern recognition accuracy of ≈94% on the Modified National Institute of Standards and Technology (MNIST) handwritten dataset when incorporated into a neural network, demonstrating its potential as an effective component in neuromorphic systems. The successful implementation of the 2D FeFET device paves the way for the development of high-efficiency, ultralow-power neuromorphic hardware which is in sub-femtojoule (48 aJ/spike) and fast response (1 µs), which is 104 folds faster than human synapse (≈10ms). The results of the research underline the potential of nanoscale ferroelectric and 2D materials in building the next generation of artificial intelligence technologies.

In this work, we proposed a three-dimensional (3-D) channel stacked array architecture based on charge-trap flash (CTF) memory for an artificial neural network accelerator. The proposed synapse array architecture could be a promising solution for implementing efficiently a large-size artificial neural network on a limited-size hardware chip. We designed a full array architecture including a stacked layer selection circuit. In addition, we investigated the synaptic characteristics of CTF device by using technology computer-aided design (TCAD) simulation. We demonstrated the feasibility of the synapse array for neural network accelerators through a system-level MATLAB simulation with the Modified National Institute of Standards and Technology (MNIST) database.

The vast application of artificial intelligence in numerous fields—image generation being one of them—has been made possible by the quick development of deep learning. Generative Adversarial Networks (GAN) can generate high-quality images through an adversarial training mechanism. The use and performance of GAN and its conditional variation, CGAN, in the field of handwritten digital image generation, are thoroughly examined in this research. The basic GAN and CGAN models, based on the PyTorch deep learning framework and the Modified National Institute of Standards and Technology (MNIST) dataset, are applied to generate handwritten digital images respectively. To assess and compare the variations between the two models concerning the fineness of image generation, the loss changes, and other relevant factors, the generation outcomes and the loss changes that occur during the training phase are documented. The experimental results demonstrate that, compared with the basic GAN, CGAN exhibits notable advantages in terms of image quality stability, the avoidance of model collapse, and the control of image categories. Furthermore, an investigation of other cutting-edge generating models indicates that there is still room for optimization in the CGAN network structure to improve its performance for increasingly intricate generative tasks.

<abstract><p>Federated learning (FL) is a framework which is used in distributed machine learning to obtain an optimal model from clients' local updates. As an efficient design in model convergence and data communication, cloud-edge-client hierarchical federated learning (HFL) attracts more attention than the typical cloud-client architecture. However, the HFL still poses threats to clients' sensitive data by analyzing the upload and download parameters. In this paper, to address information leakage effectively, we propose a novel privacy-preserving scheme based on the concept of differential privacy (DP), adding Gaussian noises to the shared parameters when uploading them to edge and cloud servers and broadcasting them to clients. Our algorithm can obtain global differential privacy with adjustable noises in the architecture. We evaluate the performance on image classification tasks. In our experiment on the Modified National Institute of Standards and Technology (MNIST) dataset, we get 91% model accuracy-layer HFL-DP, our design is more secure while as being accurate.</p></abstract>

This study introduces an innovative approach to image classification that uses Gaussian copulas with an Empirical Cumulative Distribution Function (ECDF) approach. The strategic use of distribution functions as feature descriptors simplifies the approach and enables a better understanding of the correlation structure between features in the image. This approach helps the model understand the contextual relationships between different parts of the image, resulting in a more abstract representation than a direct representation of individual pixel values. The proposed model utilizes the Distribution Function of the Distribution Value (DFDV) as the margin distribution. The Modified National Institute of Standards and Technology (MNIST) dataset is comprehensively used to assess the effectiveness of this model. The results show that the model achieves a noteworthy level of accuracy, with an average accuracy of 62.22% and a peak accuracy of 96.92%. This success was obtained by applying the Inference Function for Marginals (IFM) principles during the training stage.