Computer Memory Research Articles

Imaging swiFTly: streaming widefield Fourier Transforms for large-scale interferometry

We describe a scalable distributed imaging algorithm framework for next-generation radio telescopes, managing the Fourier transform from apertures to sky (or vice versa) with a focus on minimising memory load, data transfers, and computation. Our algorithm uses smooth window functions to isolate the influence between specific regions of spatial-frequency and image space. This allows the distribution of image data between nodes and the construction of segments of frequency space exactly when and where needed. The developed prototype distributes terabytes of image data across many nodes, while generating visibilities at throughput and accuracy competitive with existing software. Scaling is demonstrated to be better than cubic in problem complexity (for baseline length and field of view), reducing the risk involved in growing radio astronomy processing to large telescopes like the Square Kilometre Array.

Emerging 2D Ferroelectric Devices for In-Sensor and In-Memory Computing.

The quantity of sensor nodes within current computing systems is rapidly increasing in tandem with the sensing data. The presence of a bottleneck in data transmission between the sensors, computing, and memory units obstructs the system's efficiency and speed. To minimize the latency of data transmission between units, novel in-memory and in-sensor computing architectures are proposed as alternatives to the conventional von Neumann architecture, aiming for data-intensive sensing and computing applications. The integration of two-dimensional (2D) materials and 2D ferroelectric materials has been expected to build these novel sensing and computing architectures due to the dangling-bond-free surface, ultra-fast polarization flipping and ultra-low power consumption of the 2D ferroelectrics. Here, we review the recent progress of 2D ferroelectric devices for in-sensing and in-memory neuromorphic computing. Experimental and theoretical progresses on 2D ferroelectric devices, including passive ferroelectrics-integrated 2D devices and active ferroelectrics-integrated 2D devices, are reviewed followed by the integration of perception, memory, and computing application. Notably, 2D ferroelectric devices have been used to simulate synaptic weights, neuronal model functions, and neural networks for image processing. As an emerging device configuration, 2D ferroelectric devices have the potential to expand into the sensor-memory and computing integration application field, leading to new possibilities for modern electronics. This article is protected by copyright. All rights reserved.

Low-Temperature Nanosecond Laser Process of HZO-IGZO FeFETs toward Monolithic 3D System on Chip Integration.

Ferroelectric field-effect transistors (FeFETs) are increasingly important for in-memory computing and monolithic 3D (M3D) integration in system-on-chip (SoC) applications. However, the high-temperature processing required by most ferroelectric memories can lead to thermal damage to the underlying device layers, which poses significant physical limitations for 3D integration processes. To solve this problem, the study proposes using a nanosecond pulsed laser for selective annealing of hafnia-based FeFETs, enabling precise control of heat penetration depth within thin films. Sufficient thermal energy is delivered to the IGZO oxide channel and HZO ferroelectric gate oxide without causing thermal damage to the bottom layer, which has a low transition temperature (<250 °C). Using optimized laser conditions, a fast response time (<1 µs) and excellent stability (cycle >106, retention >106 s) are achieved in the ferroelectric HZO film. The resulting FeFET exhibited a wide memory window (>1.7V) with a high on/off ratio (>105). In addition, moderate ferroelectric properties (2·Pr of 14.7 µCcm-2) and pattern recognition rate-based linearity (potentiation: 1.13, depression: 1.6) are obtained. These results demonstrate compatibility in HZO FeFETs by specific laser annealing control and thin-film layer design for various structures (3D integrated, flexible) with neuromorphic applications.

Multi-Task Scenario Encrypted Traffic Classification and Parameter Analysis

The widespread use of encrypted traffic poses challenges to network management and network security. Traditional machine learning-based methods for encrypted traffic classification no longer meet the demands of management and security. The application of deep learning technology in encrypted traffic classification significantly improves the accuracy of models. This study focuses primarily on encrypted traffic classification in the fields of network analysis and network security. To address the shortcomings of existing deep learning-based encrypted traffic classification methods in terms of computational memory consumption and interpretability, we introduce a Parameter-Efficient Fine-Tuning method for efficiently tuning the parameters of an encrypted traffic classification model. Experimentation is conducted on various classification scenarios, including Tor traffic service classification and malicious traffic classification, using multiple public datasets. Fair comparisons are made with state-of-the-art deep learning model architectures. The results indicate that the proposed method significantly reduces the scale of fine-tuning parameters and computational resource usage while achieving performance comparable to that of the existing best models. Furthermore, we interpret the learning mechanism of encrypted traffic representation in the pre-training model by analyzing the parameters and structure of the model. This comparison validates the hypothesis that the model exhibits hierarchical structure, clear organization, and distinct features.

Internet of Things intrusion detection: Research and practice of NSENet and LSTM fusion models

To address the problems of complex environment, limited device computational resources and limited memory resources in the existing IoT, SELSTM, an intrusion detection system composed of NSENet and LSTM fusion based on SENet, is investigated. The NSENet part of the SELSTM system is based on the squeeze-and-excitation network (SENet). The lightweight computational modules NonLocal, SKConv and inverted residuals are fused into SE blocks, and self-attention of Nonlocal is used to improve the local receptive field of feature extraction. The channel attention and spatial attention of each part of the data are strengthened by the use of SKConv To enhance the adaptive convolution ability of the model and ensure the completeness of the information, the properties of the inverted residual structure are used to ensure that the gradient of the model decreases steadily without gradient explosion or disappearance. For the problem of data imbalance, the dataset is randomly resampled using the weight resampling technique to improve the balance of the dataset to ensure that the final detection effect of the model is more effective and generalized, while the data flow is divided into two parts for processing, and the model parameters are optimized using the model gradient optimizer consisting of the optimizer Lion and the optimization function Lookahead. The model extracts the spatial and temporal features of the data through multidimensional extraction to ensure the completeness of the data feature information in multiple dimensions, thus obtaining better detection results. The results of the experiments comparing the SELSTM model with other models on the intrusion dataset show that the intrusion detection model has a higher detection precision and accuracy than the traditional deep learning intrusion detection model, which indicates that the SELSTM has better detection performance properties and better practicality and effectiveness on IoT devices.

CAD and constructive solid geometry modeling of the Molten Salt Reactor Experiment with OpenMC

In this study, we present a detailed comparison of two independently developed models of the Molten Salt Reactor Experiment (MSRE) for Monte Carlo particle transport simulations: the constructive solid geometry (CSG) model that was developed in support of the MSRE benchmark in the International Handbook of Evaluated Reactor Physics Benchmark Experiments, and a CAD model that was developed by Copenhagen Atomics. The original Serpent reference CSG model was first converted to OpenMC’s input format so that it could be systematically compared to the CAD model, which was already available as an OpenMC model, using the same Monte Carlo code. Results from simulations using the Serpent and OpenMC CSG models showed that keff agreed within 10 pcm while the flux distribution in space and energy generally agreed within 0.1%. Larger differences were observed between the OpenMC CAD and CSG models; notably, the keff computed for the CAD model was 1.00872, which is more than 1% lower than the value for the CSG model and much closer to experiment. Several areas of the reactor that were modeled differently in the CSG and CAD models were discussed and, in several cases, their impact on keff was quantified. Lastly, we compared the computational performance and memory usage between the CAD and CSG models. Simulation of the CSG model was found to be 1.4–2.3× faster than simulation of the CAD model based on the Embree ray tracer while using 4× less memory, highlighting the need for continued improvements in the CAD-based particle transport ecosystem. Finally, major performance degradation was observed for CAD-based simulations when using the MOAB ray tracer.

Computationally Efficient and Interface Accurate Dual-Grid Phase-Field Simulation of Turbulent Drop-Laden Flows

Abstract In this work, we develop a dual-grid approach for the direct numerical simulations (DNS) of tur- bulent multiphase flows in the framework of the phase-field method (PFM). With the dual-grid approach, the solution of the Navier-Stokes equations (flow-field) and of the Cahn-Hilliard equa- tion (phase-field) are performed on two different computational grids. In particular, a base grid - fine enough to resolve the flow down to the Kolmogorov scale - is used for the solution of the Navier-Stokes equations, while a refined grid - required to improve the description of small interfacial structures - is used for the solution of the Cahn-Hilliard equation (phase-field method). The proposed approach is validated, and its computational efficiency is evaluated considering the deformation of a drop in a two-dimensional shear flow. Analyzing the computational time and memory usage, we observe a reduction between ≃30% and ≃40% (with respect to the single-grid approach), depending on the grid refinement factor employed for the phase-field variable. The applicability of the approach to a realistic three-dimensional case is also discussed, by focusing on the breakage of a thin liquid sheet inside a turbulent channel flow. Indications on the grid resolution representing a good compromise between accuracy and computational efficiency in drop-laden turbulence are also provided.

Highly Reliable Magnetic Memory-Based Physical Unclonable Functions.

Magnetic random-access memory (MRAM), which stores information through control of the magnetization direction, offers promising features as a viable nonvolatile memory alternative, including high endurance and successful large-scale commercialization. Recently, MRAM applications have extended beyond traditional memories, finding utility in emerging computing architectures such as in-memory computing and probabilistic bits. In this work, we report highly reliable MRAM-based security devices, known as physical unclonable functions (PUFs), achieved by exploiting nanoscale perpendicular magnetic tunnel junctions (MTJs). By intentionally randomizing the magnetization direction of the antiferromagnetically coupled reference layer of the MTJs, we successfully create an MRAM-PUF. The proposed PUF shows ideal uniformity and uniqueness and, in particular, maintains performance over a wide temperature range from -40 to +150 °C. Moreover, rigorous testing with more than 1584 challenge-response pairs of 64 bits each confirms resilience against machine learning attacks. These results, combined with the merits of commercialized MRAM technology, would facilitate the implementation of MRAM-PUFs.

Computational Study on Interlocked-Ferroelectricity-Contributed High-Performance Memristors Based on Two-Dimensional van der Waals Ferroelectric Semiconductors.

In order to realize the prevailing artificial intelligence technology, memristor-implemented in-memory or neuromorphic computing is highly expected to break the bottleneck of von Neumann computers. Although high-performance memristors have been vigorously developed in labs or in industry, systematic computational investigations on memristors are seldom. Hence, it is urgent to provide theoretical or computational support for the exploration of memristor operating mechanisms or the screening of memristor materials. Here, a computational method based on the main input parameters learned from the first-principles calculations was developed to measure resistance switching of two-terminal memristors with sandwiched metal/ferroelectric semiconductor/metal architectures, which strikingly agrees with the experimental measurements. Based on our developed method, the diverse multiterminal memristors were designed to fully exploit the application of interlocked ferroelectricity of a ferroelectric semiconductor and realize their heterosynaptic plasticity, and their heterosynaptic behaviors can still be well described. Our developed method can provide a paradigm for the emulation of ferroelectric memristors and inspire subsequent computational exploration. Furthermore, our study also supplies a device optimization strategy based on the interlocked ferroelectricity and easy processing of two-dimensional van der Waals ferroelectric semiconductors, and our proposed heterosynaptic memristors still await further experimental exploration.

TA-YOLO: a lightweight small object detection model based on multi-dimensional trans-attention module for remote sensing images

AbstractObject detection plays a vital role in remote sensing applications. Although object detection has achieved proud results in natural images, these methods are difficult to be directly applied to remote sensing images. Remote sensing images often have complex backgrounds and small objects, which results in a highly unbalanced distribution of foreground and complex background information. In order to solve the above problems, this paper proposes a multi-head channel and spatial trans-attention (MCSTA) module, which performs remote pixel interaction from the channel and spatial dimensions respectively to complete the attention feature capture function. It is a plug-and-play module that can be easily embedded in any other natural image object detection convolutional neural network, making it quickly applicable to remote sensing images. First, in order to reduce computational complexity and improve feature richness, we use a special linear convolution to obtain three projection features instead of the simple matrix multiplication transformation in Transformer. Second, we obtain trans-attention maps in different dimensions in a manner similar to the self-attention mechanism to capture the interrelationships of features in channels and spaces. In this process, we use a multi-head mechanism to perform parallel operations to improve speed. Furthermore, in order to avoid large-scale matrix operations, we specially designed an attention blocking mode to reduce computer memory usage and increase operation speed. Finally, we embedded the trans-attention module into YOLOv8, added a new detection head and optimized the feature fusion method, thus designing a lightweight small object detection model named TA-YOLO for remote sensing images. It has fewer parameters than the benchmark model YOLOv8, and its mAP on the PASCAL VOC and VisDrone data sets increased by 1.3% and 6.2% respectively. The experimental results prove the powerful function of the trans-attention module and the excellent performance of TA-YOLO.

Demonstration of In-Memory Biosignal Analysis: Novel High-Density and Low-Power 3D Flash Memory Array for Arrhythmia Detection.

Smart healthcare systems integrated with advanced deep neural networks enable real-time health monitoring, early disease detection, and personalized treatment. In this work, a novel 3D AND-type flash memory array with a rounded double channel for computing-in-memory (CIM) architecture to overcome the limitations of conventional smart healthcare systems: the necessity of high area and energy efficiency while maintaining high classification accuracy is proposed. The fabricated array, characterized by low-power operations and high scalability with double independent channels per floor, exhibits enhanced cell density and energy efficiency while effectively emulating the features of biological synapses. The CIM architecture leveraging the fabricated array achieves high classification accuracy (93.5%) for electrocardiogram signals, ensuring timely detection of potentially life-threatening arrhythmias. Incorporated with a simplified spike-timing-dependent plasticity learning rule, the CIM architecture is suitable for robust, area- and energy-efficient in-memory arrhythmia detection systems. This work effectively addresses the challenges of conventional smart healthcare systems, paving the way for a more refined healthcare paradigm.

Cooltools: Enabling high-resolution Hi-C analysis in Python.

Chromosome conformation capture (3C) technologies reveal the incredible complexity of genome organization. Maps of increasing size, depth, and resolution are now used to probe genome architecture across cell states, types, and organisms. Larger datasets add challenges at each step of computational analysis, from storage and memory constraints to researchers' time; however, analysis tools that meet these increased resource demands have not kept pace. Furthermore, existing tools offer limited support for customizing analysis for specific use cases or new biology. Here we introduce cooltools (https://github.com/open2c/cooltools), a suite of computational tools that enables flexible, scalable, and reproducible analysis of high-resolution contact frequency data. Cooltools leverages the widely-adopted cooler format which handles storage and access for high-resolution datasets. Cooltools provides a paired command line interface (CLI) and Python application programming interface (API), which respectively facilitate workflows on high-performance computing clusters and in interactive analysis environments. In short, cooltools enables the effective use of the latest and largest genome folding datasets.

High fidelity FEM based on deep learning for arbitrary composite material structure

Due to the outstanding performance, composite materials are widely used and analyzing their properties and designing them based on performance has become a crucial task in the field of many manufacturing industries. Composite materials possess complex multiscale structures, and traditional fine-scale finite element modeling and analysis may lead to severe computational resource challenges. To overcome this difficulty, breakthroughs in key technologies of multiscale accelerated analysis algorithms are required. In this study, an innovative approach based on artificial intelligence and multiscale finite element method is presented. This approach involves partitioning the entire composite material structure into coarse grids that resemble homogenous structures of similar size, providing results consistent to fine-grid finite element analysis. By utilizing CNN for image feature recognition and employing the CGAN adversarial method, coarse-grid equivalent stiffness matrices and multiscale shape functions from completely random microstructures of composite materials can be obtained. Consequently, this enables a rapid response process from microstructure to low-resolution grid to high-resolution physical field, with remarkably accurate physical field results. Moreover, compared to traditional fine-grid finite element methods, this approach significantly reduces memory usage and computation time. This method is applicable to composite materials with varying shaped inclusions, different component properties, and diverse geometric distributions, allowing these materials to perform high-fidelity finite element calculations on coarse grids and predict their mechanical behavior. Furthermore, this breakthrough opens avenues for accelerating the optimization design of composite materials with diverse mechanical functionalities, by employing a bottom-up approach.

Nymphaea Alba for Resistive Switching Devices: Exploring the Non‐Volatile Memory and Neuromorphic Computing Applications of the Plant Leaves

AbstractBiomaterial‐based resistive switching memory is one promising technology that could significantly impact the current direction of non‐volatile memory and neuromorphic computing. Given this, we report the fabrication of a resistive switching (RS) device based on Nymphaea Alba as an active switching layer and investigate its RS properties for memory and neuromorphic computing applications. The Nymphaea Alba switching layer was deposited by the conventional doctor blade method. Further, this is characterized by various analytical techniques. The fabricated device showed bipolar RS behavior and demonstrated good non‐volatile memory properties. Additionally, the device mimics various synaptic properties of the human brain, such as potentiation‐depression, Excitatory post‐synaptic current (EPSC), and Paired‐pulse facilitation (PPF). The charge‐flux relation reveals the non‐ideal memristor‐type behavior of the device. The Ohmic and Child's square law dominates the conduction of the device and the filamentary process responsible for the RS process of the device. These results show that the Nymphaea Alba is a promising biomaterial for non‐volatile memory applications.

Restricting Conformational Space: A New Blueprint for Electrically Switchable Self-Assembled Monolayers.

Tunnel junctions comprising self-assembled monolayers (SAMs) from liquid crystal-inspired molecules show a pronounced hysteretic current-voltage response, due to electric field-driven dipole reorientation in the SAM. This renders these junctions attractive device candidates for emerging technologies such as in-memory and neuromorphic computing. Here, the novel molecular design, device fabrication, and characterization of such resistive switching devices with a largely improved performance, compared to the previously published work are reported. Those former devices suffer from a stochastic switching behavior limiting reliability, as well as from critically small read-out currents. The present progress is based on replacing Al/AlOx with TiN as a new electrode material and as a key point, on redesigning the active molecular material making up the SAM: a previously present, flexible aliphatic moiety has been replaced by a rigid aromatic linker, thereby introducing a molecular "ratchet". This restricts the possible molecular conformations to only two major states of opposite polarity. The above measures have resulted in an increase of the current density by five orders of magnitude as well as in an ON/OFF conductance ratio which is more than ten times higher than the individual scattering ranges of the high and low resistance states.

Approaching the Ideal Linearity in Epitaxial Crystalline-Type Memristor by Controlling Filament Growth.

Brain-inspired neuromorphic computing has attracted widespread attention owing to its ability to perform parallel and energy-efficient computation. However, the synaptic weight of amorphous/polycrystalline oxide based memristor usually exhibits large nonlinear behavior with high asymmetry, which aggravates the complexity of peripheral circuit system. Controllable growth of conductive filaments is highly demanded for achieving the highly linear conductance modulation. However, the stochastic behavior of the filament growth in commonly used amorphous/polycrystalline oxide memristor makes it very challenging. Here, the epitaxially grown Hf0.5Zr0.5O2-based memristor with the linearity and symmetry approaching ideal case is reported. A layer of Cu nanoparticles is inserted into epitaxially grown Hf0.5Zr0.5O2 film to form the grain boundaries due to the breaking of the epitaxial growth. By combining with the local electric field enhancement, the growth of filament is confined in the grain boundaries due to the fact that the diffusion of oxygen vacancy in crystalline lattice is more difficult than that in the grain boundaries. Furthermore, the decimal operation and high-accuracy neural network are demonstrated by utilizing the highly linear and multi-level conductance modulation capacity. This method opens an avenue to control the filament growth for the application of resistance random access memory (RRAM) and neuromorphic computing.

Manufacturability Evaluation of Magnetic Tunnel Junction-Based Computational Random Access Memory

Manufacturability Evaluation of Magnetic Tunnel Junction-Based Computational Random Access Memory

Self-powered photodetector based on 1D TiO2-3D CdS mixed dimensional heterostructure fabricated at low temperature

Solution processed photodetectors have garnered great attention in applications such as, machine vision perception, neuromorphic computing and opto-electronic memory storage. Though, such photodetectors offer several advantages such as ease of fabrication, high scalability, low thermal budget and low-cost processing, multi-modal functionality etc. however, they suffer from the major drawback of inferior device performance −as low responsivity and slow rise time, particularly due to the intrinsic poor crystallinity of the photoactive material. In this work, we demonstrate a solution processed photodetector with impressive performance at comparatively low processing temperatures (<150 °C) based on the mixed dimensional heterostructure configuration of 1D TiO2 nanorods and 3D CdS nanoflowers. TiO2 nanorods have been synthesized by hydrothermal technique, whereas their CdS sensitization is done by chemical bath deposition. Low cost carbon paste is used as electrode instead of conventional non-economic noble metal electrodes. X-ray diffraction studies validated excellent crystallinity of the photoactive material even under low temperature processing condition. The type-II Heterojunction (TiO2 and CdS) configuration photodetector shows efficient response at zero bias, thus yielding a self-powered device. The detector shows response in UV and visible region, with excellent responsivity of 110 mA/W (5 V), 563 µA/W (0 V) and a quicker rise time of 81 ms. Albeit the simple fabrication scheme and low processing temperatures, the detector exhibited promising figures-of-merit, which aids in fabrication of novel solution processed photodetectors.

GA-GGD: Improving semantic discriminability in graph contrastive learning via Generative Adversarial Network

Graph contrastive learning has garnered considerable research interest due to its ability to effectively embed graph data without manual labels. Among them, methods based on Deep Graph Infomax (DGI) have been widely studied and favored in the industry because of their fast training speed, applicability to large-scale data. DGI-based methods usually obtain a noise graph through node shuffling. The proxy task of these methods encourages the encoder to distinguish whether the nodes come from the original graph or the noise graph, thereby maximizing the mutual information between the node representation and the graph it belongs to, while also maximizing the Jenson-Shannon divergence between the nodes of original graph and noise graph. However, we argue that these approaches only enable the encoder to differentiate between semantically meaningful graphs and noise graphs, but not to effectively identify different semantic graphs. This leads to the inability of the encoder to effectively embed information between different semantics, significantly reducing the robustness and affecting the performance of downstream tasks. In addition, this training mode makes the model more sensitive to attacks. To improve their semantic discriminability, we take advantage of the natural ability of generative adversarial networks to generate semantic data, proposing a method called Generative Adversarial Graph Group Discrimination (GA-GGD). Specifically, it consists of a graph group discriminator and a semantic attack generator. The discriminator aims to encode the graph and identify whether nodes originate from the original graph. The goal of the generator is to use random features and graph structure to find vulnerabilities of discriminator and generate node representations with similar but wrong semantic to confuse the discriminator. GA-GGD can improve the model’s semantic information embedding without significantly increasing computational overhead and memory occupancy. We test the effectiveness of the proposed model on commonly used data sets and large-scale datasets, as well as in various downstream tasks such as classification, clustering, and adversarial attacks defence. A wealth of experimental results confirm the efficacy of the proposed model.

Low Power Content Addressable Memory Designing and Implementation Using Voltage Swing Self Adjustable Match Line Technique

One of the essential components of computer systems is memory. A primary hindrance in this regard is the memory speed. Content Addressable Memory (CAM) speeds up transformations and table lookups in network routers and data processing systems for hardware search engines. Parallel seeks using the CAM (Content Addressable Memory) model are often used to enhance memory performance. This paper uses the voltage swing self-adjustable match line (VSSA-ML) technique to describe low-power content addressable memory design and implementation. This project decreases Match Line (ML) power loss by reducing load capacitance and ML voltage swing. A simple ML voltage detector is proposed instead of the complex, fully different detector that allows ML voltage swings near zero. This paper presents 6T 8×8 CAM arrays using VSSA-ML Technique using Tanner tools 45-nm technology. On the other hand, this design enhances robustness in processing variations by self-adjusting voltage swings. Implementation analysis states that the described mode 6T 8×8 CAM design utilized fewer MOSFETs than the 8T 8×8 CAM array.