Memory Address Research Articles (Page 3)

Emerging non-volatile memories (eNVM) promise to solve the imminent von Neumann bottleneck by enabling future computing systems to utilize the computing-in-memory (CIM) paradigm offering exceptional energy efficiency and performance advantages. As Moore's law becomes obsolete, CIM architectures are prominent candidates to push the boundaries of existing computing systems and usher in a new generation of computing models, such as neuromorphic systems. Furthermore, conventional systems face another significant problem in addition to the von Neumann bottleneck. Hardware security threats (e.g., Rowhammer) have gained momentum and can expose an entirely pristine attack surface for adversaries. These vulnerabilities distinguish themselves by being particularly challenging to patch because their origin lies in the rigid hardware layout. Unfortunately, neuromorphic systems are no exception. We presented NeuroHammer as one of the first unique hardware security attacks on eNVMs, enabling an attacker to intentionally flip bits in memristive crossbar arrays. This article extends our previous results by thoroughly examining the underlying concepts leading to the NeuroHammer attack. First, we investigate memory access patterns to gain insight into the tangible impact of NeuroHammer. Second, we extend our simulation methodology to accommodate transistor/one resistive (1T1R) crossbar structures and prove the prevalence of the NeuroHammer attack. Finally, we discuss the real-world implications of NeuroHammer on CIM architectures.

This paper presents a new epidemiological modeling approach that adapts the working set (WS) concept from computer memory management to the dynamics of infectious diseases. Traditional compartmental models provide valuable insights but are limited in their ability to capture dynamic isolation and heterogeneous contact patterns. In contrast, the WS model conceptualizes a time-varying subset of agents actively participating in social interactions, allowing for dynamic adjustments to the rate of infection and the explicit identification of superspreaders. By incorporating isolation states for both susceptible and infected individuals, the model more realistically captures quarantine and targeted interventions. Including an incubation period reduces epidemic peaks by nearly 40% and delays them by more than three weeks, providing critical time for public health response. Within the WS model, moderate isolation reduces peak infection rates by more than three times compared to uncontrolled scenarios, while high isolation almost completely prevents large-scale spread. These results highlight the model's ability to estimate the intensity and timing of interventions with greater accuracy than traditional models. By integrating the time window parameter and computer resource management principles, the adapted WS model represents a robust and adaptable tool for analyzing epidemic dynamics. The results highlight its potential for advancing epidemic modeling and supporting real-time public health decision-making.

Context . Cosmic voids are key elements in our understanding of the large-scale structure of the Universe. They are crucial for constraining cosmological parameters, understanding the structure formation, and evolution of our Universe, and they could also serve as pristine laboratories for studying galaxy formation without all the hassle due to environmental effects. Thus, the ability to accurately and consistently identify voids, both in numerical simulations and in observations, is essential. Aims . We present the Algorithm for Void Identification in coSMology ( AVISM ), a new void finder for analysing both cosmological simulation outputs and observational galaxy catalogues. In the first case, the code handles raw particle or cell data, dark matter haloes, and synthetic galaxy catalogues. For observational data, the code should be coupled with external tools that provide the required dynamical information to apply the algorithm. This new numerical tool is efficient in terms of computational resources, both wall time and memory. Methods . A set of numerical tests designed to assess the code’s capabilities were carried out, including parameter robustness, computational performance, and the use of different matter components in a cosmological simulation. AVISM’s performance was also compared, both statistically and on a one-to-one basis, with the state-of-the-art void finders DIVE and ZOBOV using a dark matter halo catalogue from a large-volume cosmological simulation. An application to a galaxy survey is also provided to demonstrate the code’s ability to handle real data. Results . We designed a new void finder algorithm that combines geometrical and dynamical information to identify void regions and a hierarchical merging process to reconstruct the entire 3D structure of the void. The outcome of this process is a void catalogue with complex boundaries without assuming a prior shape. This process can be repeated at different levels of resolution using finer grids, leading to a list of voids-in-voids and a proper description of void substructure. Conclusions . We present and release AVISM , a new publicly available void finder.

While Vision Transformers (ViTs) have recently demonstrated impressive performance in computer vision tasks, their high computational demands and memory usage limit their applicability in real-time and edge AI scenarios. In contrast, Convolutional Neural Networks (CNNs) remain the preferred choice for such environments due to their lower latency, inductive bias, and efficiency. This study examines the impact of five widely used convolutional operations (standard 2D spatial, grouped, shuffle, depthwise separable and shift) when integrated into the ResNet-50 architecture. All model variants are trained on the Tiny-ImageNet and CIFAR-10/100 dataset under standardized GPU-based settings and evaluated across three edge AI platforms: Raspberry Pi 5, Coral Dev Board and Jetson Nano. The analysis includes parameter count, FLOPs, accuracy, inference time, power consumption and detailed runtime decomposition. Results show that while depthwise separable convolutions offer theoretical efficiency, they suffer from increased memory access on memory-bound platforms. In contrast, shuffle and shift convolutions yield better trade-offs between accuracy, computational load, and inference speed. These findings provide actionable insights for designing hardware-aware, deployment-optimized CNN architectures suitable for resource-constrained applications.

In the era of artificial intelligence, Internet of things and big data, processing massive data puts forward unprecedented requirements for the throughput and energy efficiency of computing systems. In traditional von Neumann architectures, frequent data movement between processor and memory results in significant energy consumption and latency, known as theMemory Wall problem. In this paper, the principle, application and performance of variable resistance random-access memory (RRAM), magnetoresistive random-access memory (MRAM) and ferroelectric random-access memory (Feram) in the memory computing (CIM) structure are studied in depth. CIM technology is considered to be the key to overcome theMemory wall bottleneck inherent in the traditional von Neumann architecture. Firstly, the physical mechanism and characteristics of these three storage technologies are systematically described. Subsequently, a detailed analysis of their effectiveness in various application scenarios is provided through innovative simulation designs: RRAM-based neuromorphic chips (neurrams) exhibit superior energy efficiency in simulation calculations; MRAM exhibits performance close to conventional memory in non-volatile caching and in-memory logic applications; while FeRAM has unique advantages in ultra-low-power binary neural networks (BNNS). A comprehensive comparative analysis demonstrates the complementarity of their technical paths, and proposes future integration strategies for heterogeneous computing systems. This study adopts a co-design perspective across device, circuit, and architecture levels to provide important theoretical foundations and design insights for the development of next-generation efficient heterogeneous computing systems.

Currently, computational analysis of vibration-induced noise in electromagnetic equipment cores predominantly relies on three-dimensional finite element simulation methods. However, this approach frequently encounters challenges when addressing complex nonlinear magnetic characteristics and elastic mechanical problems, including substantial computational scale, protracted solution times, and stringent demands on computer memory resources. Consequently, it struggles to accommodate the rapid iteration and computational requirements inherent in the engineering design. To address this challenge, this paper establishes an electromagnetic–mechanical–acoustic multiphysics coupled numerical model for magnetically saturated controllable reactor cores based on electromagnetic coupling and elastic mechanics principles. Through computational simulations, it obtains key parameters including winding current, core magnetic flux density, vibration displacement, and noise characteristics under various operating conditions. Concurrently, experimental data acquisition is conducted to validate the model's effectiveness by comparing simulation results with experimental measurements. Results demonstrate excellent agreement between measured, simulated, and calculated values for excitation current, core magnetic flux density, vibration displacement, and noise distribution. Furthermore, compared to the three-dimensional finite element model, the proposed model achieves a computational speed improvement of 310–350 times, with storage space requirements reduced to between 1/6800 and 1/7000 of the original model. This significantly enhances efficiency while maintaining computational accuracy. The validity and accuracy of this method have been verified, providing an efficient theoretical basis and computational approach for rapidly calculating the magnitude and distribution of reactor vibration and noise during the design phase, as well as exploring novel noise reduction measures.

Reduced order modeling is of paramount importance for efficiently inferring high-dimensional spatio-temporal fields in parametric contexts. However, conventional dimensionality reduction techniques are typically limited to known and constant parameters, inefficient for nonlinear and chaotic dynamics, and uninformed to the actual system behavior. In this work, we propose a SHallow REcurrent Decoder-based Reduced Order Modeling technique (SHRED-ROM) capable of reconstructing high-dimensional state dynamics in multiple scenarios from the temporal history of limited sensor measurements. To enhance computational efficiency and memory usage, we reduce data dimensionality through data- or physics-driven basis expansions, allowing for compressive training of lightweight networks with minimal hyperparameter tuning. Through applications on chaotic and nonlinear fluid dynamics, we show that SHRED-ROM is a robust decoding-only strategy, capable of dealing with both fixed or mobile sensors, physical and geometrical (possibly time-dependent) parametric dependencies and different data sources, such as high-fidelity simulations, coupled fields and videos, while being agnostic to sensor placement and parameter values.

Abstract Ferroelectric capacitors (Fe‐Caps) exhibiting memcapacitive behavior are emerging as a compelling platform for low‐power capacitive in‐memory computing (IMC). Their inherently open‐circuit nature suppresses static current and sneak paths, offering a distinct advantage over resistive IMC counterparts. In this work, an experimental demonstration is presented of a cross‐point Fe‐Cap array prototype, establishing capacitive IMC with non‐destructive readout characteristics. The devices exhibit robust reliability, sustaining more than 10 5 read cycles and 10 7 write cycles, with retention projected beyond 10 years without memory collapse. At the system level, a 4 × 2 array achieves 100% accuracy in 2‐pixel × 2‐pixel pattern‐recognition tasks, while scaling to a 16 × 4 array and letter‐recognition tasks highlights the outstanding scalability of Fe‐Cap‐based IMC platforms. These results not only validate the Fe‐Caps as both a functional and reliable IMC element but also provide an experimental framework for straightforward evaluation of IMC functionality, marking a milestone toward practical and energy‐efficient neuromorphic hardware.

Memristors are poised to revolutionize neuromorphic computing and next-generation memory, yet persistent challenges, including stochastic conductance control, limited endurance, and poor scalability, impede their practical adoption. Here, we report a self-rectifying memristor leveraging a GaOx/InOx heterostructure to overcome these barriers. The device demonstrates an ultrahigh rectification ratio and nonlinearity (>107), effectively suppressing sneak currents in passive crossbar arrays. Robust multi-level conductance modulation and cycling stability exceeding 106 cycles are achieved, alongside minimal inter-device variability (σ/μ < 10%) in a 6 × 6 passive array. Critically, it demonstrates dynamic tunability between short-term and long-term plasticity, modulated via pulse number/amplitude, a cornerstone for biologically inspired learning. By implementing the one-bitline pull-up scheme, the design achieves an integration density of 125.1 Tb at a 10% read margin, outperforming state-of-the-art resistive memories. This work establishes oxide heterostructures as a scalable platform for high-density neuromorphic systems, bridging the gap between functional material innovation and hardware-driven machine intelligence.

This study proposes the word vector convolutional time-series Network (WVCTS-Net) for extracting gas feature signatures from pickled vegetables to distinguish geographical origins. First, treating electronic nose sensor responses as "word sequences" akin to NLP inputs, we introduce a word vector convolutional (WVC) block to enhance interpretability. Second, a time-domain analysis module based on gated recurrent units was designed to capture long-sequence contextual dependencies. The final WVCTS-Net integrates these components into a low-complexity gas classification framework. Compared to both generic and gas-specific models, WVCTS-Net achieves optimal performance with minimal parameters while maintaining cross-gas applicability. Structural optimizations yield a lightweight model that substantially reduces computational complexity and memory usage, demonstrating exceptional efficiency on edge devices. This work provides a robust theoretical foundation for real-time intelligent gas recognition and analysis at the terminal level.

EAAR: Efficient and Accurate Action Recognition model with enhanced spatio-temporal perception.

Positron Emission Tomography (PET) is an important clinical imaging tool but inevitably introduces radiation exposure to patients and healthcare providers. Reducing the tracer injection dose and eliminating the CT acquisition for attenuation correction can reduce the overall radiation dose, but often results in PET with high noise and bias. Thus, it is desirable to develop 3D methods to translate the non-attenuation-corrected low-dose PET (NAC-LDPET) into attenuation-corrected standard-dose PET (AC-SDPET). Recently, diffusion models have emerged as a new state-of-the-art deep learning method for image-to-image translation, better than traditional CNN-based methods. However, due to the high computation cost and memory burden, it is largely limited to 2D applications. To address these challenges, we developed a novel 2.5D Multi-view Averaging Diffusion Model (MADM) for 3D image-to-image translation with application on NAC-LDPET to AC-SDPET translation. Specifically, MADM employs separate diffusion models for axial, coronal, and sagittal views, whose outputs are averaged in each sampling step to ensure the 3D generation quality from multiple views. To accelerate the 3D sampling process, we also proposed a strategy to use the CNN-based 3D generation as a prior for the diffusion model. Our experimental results on human patient studies suggested that MADM can generate high-quality 3D translation images, outperforming previous CNN-based and Diffusion-based baseline methods. The code is available at https://github.com/tianqic/MADM.

The artificial neuronal architecture in the form of computational memories is unfolding, obliterating the conventional von Neumann computers, indicating the increase in computing speed and cost. Memristors, with their neuron-mimicking capabilities, have become an important fundamental component in neuronal logic circuits. Transition metal oxides such as VO2 undergo phase change due to the metal-to-insulator transition from tetragonal (R) to monoclinic (M1) when the temperature falls below 340 K. VO2 thin films were grown using pulsed laser deposition, and their switching behavior was observed under different voltage biases. The VO2 thin films were illuminated with laser sources at wavelengths of 405 and 980 nm. A spiking time-dependent analysis was conducted under these various illumination conditions. The results demonstrate the potential for optical-based computing. We have successfully shown that VO2 can be used as a memristor for neuronal logic applications. The ratio of the high-resistance state (RHRS) to the low-resistance state (RLRS) was found to be 6.3 × 103.

Minerals bearing rare earth elements (REEs) are formed through long geological processes, among which monazite, bastnasite, xenotime, and ionic adsorption clays are the most economically exploited. Although Brazil has one of the largest reserves of REEs on the planet, its production is still not significant on the world stage. China remains dominant, with the largest reserves of REEs and controlling more than half of world production. Due to their important application in advanced clean and low-carbon energy technologies, REEs have become fundamental to the energy transition process. Technological applications related to catalyst synthesis, ceramics production, and metallurgy have been explored. Furthermore, the use of REEs in devices of great demand today, such as computer memory, rechargeable batteries, and mobile phones, has been cited. With the growing demand for these critical minerals, large mining companies are seeking to implement cleaner production policies in their processes and save natural resources to minimize the environmental impacts of the exploration. Robust analytical techniques have made it possible to characterize these elements in multi-element geological matrices, with the increasing exploration and identification of new REE mineral reserves.

SRAM is widely used in computing-in-memory (CIM) neural network accelerators because of its relatively mature technology and good compatibility with complementary metal oxide semiconductor logic process. Digital SRAM-CIM is favored by researchers because of its stability and accuracy. However, the current digital SRAM-CIM macro only supports the weight-stationary dataflows, which means the repeated movement of graph data. Some special deep neural network layers, such as depth-wise, make the utilization of computing resources inside CIM low. To overcome these problems, we propose C-CIM, which can switch between input-stationary and weight-stationary dataflows and support matrix multiplication as well as convolution operations with multiple mainstream convolution kernel sizes (1×1, 3×3, 5×5 and 7×7). The C-CIM achieves an average performance of 27.31TOPS/W@8b at a frequency of 1GHz. Experimental results show that our proposed SRAM-CIM successfully outperforms baseline in terms of performance optimization, achieving up to 7.6× performance speedup and up to 86.84% reduction in activation relocation.

Virtual and Extended Reality technologies are increasingly adopted in fields such as healthcare, entertainment, and education. These applications heavily rely on Smart Eye-Wears (SEWs) and Artificial Intelligence (AI) to provide users with new ways to perceive their environment. However, SEWs face limitations in computational power, memory, and battery life. Offloading computations to external servers is a prominent example of edge computation. However, this also presents considerable challenges due to delays caused by varying network conditions and server workloads. This paper proposes self-adaptive techniques based on tabular reinforcement learning (RL) to optimize the offloading of Deep Neural Network tasks between the SEW, the user’s smartphone, and cloud servers. The goal is to maintain a high-quality user experience while minimizing energy consumption and 5G connection costs. We evaluated our framework under varying 5G and WiFi bandwidths and cloud latency. The results show that Q-learning, SARSA, and Expected SARSA achieve near-optimal policies, with Q-learning demonstrating superior performance in reducing execution time violations (approximately at 10%) and improving agent stability. Additionally, our approach offers a more favorable trade-off between energy efficiency and execution time violations compared to two baseline methods. Real-system experiments reveal that the proposed solution can double SEW battery life with respect to local computation while maintaining a good quality of service, with only 11% execution time violations. These findings highlight the effectiveness of our approach in managing resources and enhancing the overall user experience in SEW AI applications.

Van der Waals (vdW) layered α-In₂Se₃ possesses stable ferroelectricity even down to monolayer, showing great promise in emerging ferroelectric semiconductor non-volatile memory and in-memory computing. Deciphering the atomistic mechanisms governing its complex phase transition occurring in devices, such as thermal budget during fabrication or cycling, is of critical importance for device performance improvement yet remains challenging due to the intricate polymorphism and low interchange barriers. Here, we directly visualize the unique atomic-level in-plane directional phase transition in 2H-α In₂Se₃ that initiates from the In-Se octahedral framework, and is driven by the migrations of indium ions (in octahedrons) and vacancies (in vdW-gap), revealing a novel intralayer and interlayer indium atoms and vacancies rearrangement pathway. We demonstrate that each In-Se octahedral configuration evolves into two non-layered In-Se tetrahedral frameworks, ultimately coalescing a novel non-layered 6H-type In₂Se₃ phase with the cation sites might being occupied by 1/3 vacancy and 2/3 indium. Our results provide detailed microscopic insights on the phase transition dynamics of the ferroelectric 2H-α In₂Se₃ in response to the thermal stimulus, and may offer guidelines for the precise controlling of specific In₂Se₃ phases and the reliability improvement of ferroelectric In₂Se₃-based nanodevices.

Time delays increase the effective dimensionality of reservoirs, thus suggesting that time delays in reservoirs can enhance their performance, particularly their memory and prediction abilities. We find new closed-form expressions for memory and prediction functions of linear time-delayed reservoirs in terms of the power spectrum of the input and the reservoir transfer function. We confirm this relationship numerically for some time-delayed reservoirs using simulations, including when the reservoir can be linearized but is actually nonlinear. Finally, we use these closed-form formulas to address the utility of multiple time delays in linear reservoirs in order to perform memory and prediction, finding similar results to previous work on nonlinear reservoirs. We hope that these closed-form formulas can be used to understand memory and predictive capabilities in time-delayed reservoirs.

ResGSNet: Enhanced local attention with Global Scoring Mechanism for the early detection and treatment of Alzheimer's Disease.

A novel lightweight hand gesture recognition approach that is based on Frequency-Modulated Continuous Wave (FMCW) RADAR, which aims to minimize computational complexity and memory usage as well as maintain a high recognition performance, is proposed in this paper. Most of the existing methods utilize two-dimensional (2D) or three-dimensional (3D) features that are combined with complex neural network structures, which result in high computational costs. The proposed approach in contrast extracts four components, which include range, Doppler, azimuth, and elevation, as the one-dimensional (1D) time-series features. These features are fed into a neural network that comprises of a multi-branch temporal convolutional network, depthwise separable convolutions, and a channel attention mechanism in order to enhance the classification performance. The experiments were conducted with 9 hand gestures that were collected from 9 participants. The proposed system achieved a high accuracy of 99.38% with only 44.6K parameters and 1.84M FLOPs. Extensive ablation studies and comparative experiments against the existing models demonstrated that the proposed method effectively balances the performance and computational efficiency. This study validates the expressive capability of 1D features for hand gesture recognition and suggests practical applicability in resource-constrained environments, such as embedded systems.