The SystemC language, with its higher level of abstraction, plays a critical role in facilitating hardware/software co-design and architecture exploration. However, as most hardware models are predominantly written in Verilog and translating between SystemC and Verilog remains a challenge, an efficient and reliable tool for translating between these two languages is essential to streamline system development. This article proposes SCAV, a bidirectional translator between SystemC and Verilog, which breaks these limitations. SCAV provides a fully automated solution for translating both SystemC to Verilog and Verilog to SystemC, leveraging a translation framework with front-end/back-end separation. Additionally, SCAV incorporates an Abstract Syntax Tree (AST) filter, optimizing the translation process by filtering out invalid content. The experimental results demonstrate that SCAV achieves a 100% adaptation rate for Verilog and a 98% adaptation rate for SystemC, with 100% accuracy in both directions. Furthermore, SCAV outperforms existing tools, delivering a minimum speedup of 18% across various test cases.

This paper investigates new computer architectures for the hardware implementation of dynamic (spiking) neural networks capable to replace up-to-date networks built on neurons with a static activation function. We propose for the first time the use of a recently developed compact analog model of a spiking neuron, consisting of only three elements (a volatile memristor, a tunnel diode, and a capacitor), as the basic element of a reservoir computer of Liquid State Machine (LSM) type. A computer model of the reservoir is proposed, including 7,480 neurons and approximately 254,000 connections, with a topology formed using the biologically motivated LSM stochastic synapse distribution algorithm. The results of the proposed solution are demonstrated on the task of recognizing handwritten digits from the MNIST dataset. A classification accuracy of 93 % is achieved, which is comparable to known LSM implementations. Estimates for the proposed reservoir performance of the future hardware implementation exceed those of existing analogs by an order, and in terms of energy efficiency by 3-4 orders. Thus, the proposed study demonstrates for the first time the practical applicability of the three-element neuron model for machine learning tasks and confirms its potential as a basic element for constructing scalable and energy-efficient neuromorphic computing systems.

Abstract Recent efforts to improve the efficiency of neuromorphic and machine learning systems have centred on developing of specialised hardware for neural networks. These systems typically feature architectures that go beyond the von Neumann model employed in general-purpose hardware such as GPUs, offering potential efficiency and performance gains. However, neural networks developed for specialised hardware must consider its specific characteristics. This requires novel training algorithms and accurate hardware models, since they cannot be abstracted as a general-purpose computing platform. In this work, we present a bottom-up approach to training neural networks for hardware-based spiking neurons and synapses, built using ferroelectric capacitors (FeCAPs) and resistive random-access memories (RRAMs), respectively. Unlike the common approach of designing hardware to fit abstract neuron or synapse models, we start with compact models of the physical device to model the computational primitives. Based on these models, we have developed a training algorithm (BRUNO) that can reliably train the networks, even when applying hardware limitations, such as stochasticity or low bit precision. We analyse and compare BRUNO with Backpropagation Through Time. We test it on different spatio-temporal datasets. First on a music prediction dataset, where a network composed of ferroelectric leaky integrate-and-fire (FeLIF) neurons is used to predict at each time step the next musical note that should be played. The second dataset consists on the classification of the Braille letters using a network composed of quantised RRAM synapses and FeLIF neurons. The performance of this network is then compared with that of networks composed of LIF neurons. Experimental results show the potential advantages of using BRUNO by reducing the time and memory required to detect spatio-temporal patterns with quantised synapses.

Evaluating the resilience of sophisticated cryptographic hardware against fault injection attacks necessitates high-speed simulation environments. At the Electronic System Level (ESL), SystemC provides a valuable framework for constructing fast functional models. However, traditional fault injection techniques require direct and intrusive code modifications to these models, which complicates the security assessment process. This paper proposes a novel methodology that circumvents this limitation by leveraging Aspect-Oriented Programming (AOP). A non-intrusive fault injection and detection environment is introduced, where faults are woven into SystemC cryptographic models using AspectC++, eliminating the need for source code alterations and offering a practical alternative to complex physical cryptanalysis. This approach is validated through a thorough case study on a SystemC model of the LED lightweight algorithm, focusing on two critical aspects: the efficacy of AOP for accurate fault detection and its overhead regarding simulation performance and executable size. The results demonstrate that the proposed method effectively evaluates a design's efficiency against fault attacks. It is also confirmed that AOP integration imposes a negligible impact on simulation time, preserving the speed advantages of ESL simulation while enabling seamless security analysis.

Resistive crossbars using non-volatile memory devices have become promising components for implementing deep learning models in hardware. However, crossbar-based computations encounter a notable challenge due to device and circuit-level non-idealities. In this study, we explore three recent crossbar simulation tools, namely, AIHWKIT, CrossSim, and MemTorch, and evaluate four pre-trained convolutional neural networks (CNNs) for image classification on CIFAR-100, MNIST, and SVHN datasets using these tools. To the best of our knowledge, this is the first study using all three simulation tools of analog accelerators, representing three distinctive hardware settings, to evaluate CNN-based model robustness under analog noise and nonlinearities. We first test the robustness of four models (VGG19, InceptionV3, ResNet50, & MobilenetV2) using different levels of white Gaussian noise as baselines. Then we evaluate their inference performance on the three tools to determine their resilience to analog noise and nonlinearities in the hardware environment. Results show that while all CNNs' suffer performance degradation as expected, insights are obtained such as ResNet50 outperforms others in two out of three simulators despite real-world hardware imperfections due to its deep structure. Furthermore, we explore the impact of three key factors: cell bits, ADC resolution, and tile size on CNN performance across various hardware configurations, providing recommendations for the efficient deployment of CNN on hardware. Our results suggest that increasing cell bits and ADC resolution improves CNNs' performance. It is also observed that smaller tiles are suitable for lightweight models, while for complex networks, we should choose larger shape of tiles.

FedFreeze: A dual-phase layer freezing framework for federated learning

Learning shapes the human brain, yet structural changes underlying this process remain difficult to characterize in vivo. Recent advances in magnetic resonance imaging (MRI)-including relaxometry, magnetization transfer, proton density, and diffusion imaging-combined with improved hardware and biophysical models, now allow highly specific assessment of subtle microstructural changes during learning. Here, we review studies documenting learning-induced changes in brain microstructure. Short training intervals elicit rapid MRI-detectable changes, including increases in restricted diffusion and local tissue volume, particularly in the hippocampus, potentially reflecting early neurite and glial adaptations. Longer training periods reveal additional changes in task-relevant gray and white matter, suggestive of adaptations in myelin, neurites, and neuroglia. The link between MRI changes and behavioral improvements is inconsistent, likely due to heterogeneous temporal dynamics of plasticity and interindividual variability. Because MRI provides only indirect insight into tissue microstructure, initial studies combine complementary contrasts with multivariate statistics to reduce interpretational ambiguities. High-field imaging, cross-modal approaches such as transcranial magnetic stimulation, and cross-species studies further bridge animal models and human research. Together, these developments refine biologically grounded models of human plasticity and hold promise for translational applications in personalized learning and rehabilitation.

Pinching-antenna systems (PASSs) offer a low-complexity and reconfigurable solution for near-field downlink communications by deploying multiple radiating elements along a single waveguide. Existing studies mainly assume simplified propagation conditions or focus on spectral efficiency, while the impact of environment-dependent interference patterns arising from user-specific blockage conditions on energy-efficient design remains unclear. An energy-efficient downlink design for single-waveguide PASS based on environment-division multiple access (EDMA) is investigated. Under a given propagation environment, EDMA exploits user-dependent blockage and visibility differences through proper pinching-antenna placement, thereby inducing different multi-user interference patterns without increasing radio-frequency hardware complexity. We examine how such blockage-dependent interference influences the relationship between spectral efficiency and energy efficiency, and develop an energy-aware EDMA framework that jointly considers pinching-antenna locations and transmit power allocation under quality-of-service constraints. The resulting coupled design problem is solved through an alternating optimization procedure. EDMA is compared with conventional time-division multiple access (TDMA) using a unified hardware and power-consumption model. Numerical results reveal clear energy-efficiency threshold behaviors with respect to blockage intensity, user population, and service requirements. The results further show that EDMA can significantly outperform TDMA in specific operating regimes.

ARP spoofing continues to be a significant threat in local area networks because of the lack of inherent authentication, which permits man-in-the-middle and denial-of-service attacks. Current mitigation strategies frequently depend on specific hardware or resource-heavy machine learning models, restricting their feasibility for real-time applications. This document introduces NetShield, a lightweight system for intrusion detection and response that detects ARP spoofing and volumetric DDoS attacks through rule-based IP–MAC checks and protocol-specific traffic evaluation. NetShield actively observes live traffic via libpcap, identifies MAC discrepancies within a time frame, and automatically prevents verified attackers through firewall enforcement. A logging module facilitates the transmission of structured alerts to Firebase for immediate dashboard visualization. Test results show elevated detection precision, reduced response time, and minor system resource use, rendering NetShield ideal for small-to-medium business and educational networks

We present a comprehensive study of a tumor-immune interaction model with delayed immune activation, combining analytical, numerical, and experimental approaches. A central feature of our formulation is the use of a generalized Hill function for immune activation, where the exponent m introduces tunable cooperativity. This generalization extends beyond conventional Michaelis-Menten or fixed-saturation forms and captures a wider range of nonlinear immune responses. On the analytical side, we derive explicit conditions for transcritical and Hopf bifurcations, clarifying the roles of key biological parameters and the immune response delay in shaping tumor dynamics. Numerically, we investigate the time-series, phase-plane plots, and both single- and two-parameter bifurcation scenarios with respect to the delay and other system parameters that confirmed the observed transitions are in excellent agreement with the analytical predictions. Most importantly, we implement the delayed tumor-immune model in an analog electronic circuit by reformulating the Hill activation function in terms of hyperbolic tangents, enabling direct laboratory exploration of the system. Experimental investigations reveal steady-state and oscillatory behaviors, dependent on time delay, that closely match the numerical simulations, despite the unavoidable real-world effects such as parameter mismatch, noise, and fluctuations. To the best of our knowledge, this is the first realization of a time-delayed tumor-immune model in hardware in the physics laboratory. This provides a novel bridge between mathematical theory, numerical analysis, and experimental validation and opens new directions for probing tumor dormancy, immune oscillations, and relapse under controlled physical conditions.

Background The current stage of Kazakhstan’s development is characterized by the formation of a digital economy. One of its aspects involves improving the quality of education through its digital transformation. This approach aligns with modern global trends that facilitate the active development and operation of software, hardware, and robotic systems for organizing, maintaining, monitoring, and managing the educational process. Consequently, the objective of this study is to develop a concept for an intelligent robotic assistant to aid in conducting school lessons. The object of the study is physics lessons conducted in the primary and secondary schools of Kazakhstan until the completion of training. Methods For the digital transformation of the research object, methods of system analysis, graphical methods, simulation modeling, and object-oriented design were employed. Formal models of the application features of these methods were created, describing the input and output data, as well as their role within the digital assistant concept. Results The digital assistant concept comprises a hardware model and an autonomous robot, detailing its core functionality and its interaction with all participants in the educational process. The robot’s social functions have been defined: demonstrating educational material, answering student questions, and monitoring student behavior and classroom conditions. The structure of the educational content, utilized in the form of technological maps, has been established. Based on this, a digital twin system has been developed, implementing a scenario for the functioning of the intelligent assistant during a lesson in a virtual environment. Conclusion The developed concept provides a unified approach to teaching school children, creating conditions for the formation of individual educational trajectories and a favorable, comfortable learning environment. The obtained results constitute a comprehensive educational model that establishes a unified digital ecosystem for the educational organization.

BackgroundPost-traumatic knee replacement (PTKR) is frequently complicated by the presence of retained metallic hardware around the joint, which limits the use of intramedullary alignment guides. Consequently, extramedullary jigs are often required, although they may increase radiation exposure and reduce alignment precision. Patient-specific guides (PSGs), generated from medical imaging and produced via 3D printing, offer a potential alternative for improving accuracy in complex surgical scenarios. This study aimed to assess the accuracy of PSGs in PTKR using in-vitro knee models with and without retained hardware.MethodsCT images of arthritic knees were used to generate 3D-printed anatomical models. Metallic plates and screws were subsequently mounted to replicate typical post-traumatic hardware configurations. These phantoms underwent CT scanning for virtual surgical planning, and patient-specific guides (PSGs) were designed based on the reconstructed preoperative models. In-vitro distal femoral and proximal tibial resections were then performed by a surgeon using the corresponding PSGs. After the simulated procedures, all phantoms were re-scanned to quantify PSG positioning accuracy and resection angles.ResultsKnee phantoms with hardware exhibited shape deviations 17–18.5 times greater than those without hardware (p < 0.05). PSG positioning errors averaged 0.68 mm and 2.83° in hardware models, compared to 0.55 mm and 1.32° in non-hardware models. Resection angle errors in hardware phantoms ranged from 2.4° to 3.1°, significantly higher than in the non-hardware group.ConclusionsBased on the in-vitro experimental findings, PSGs allow PTKR to be performed without the removal of retained hardware while achieving accuracy that exceeds that of traditional extramedullary alignment techniques. Although hardware presence results in a quantifiable reduction in accuracy, PSGs continue to demonstrate improved alignment precision and contribute to enhanced workflow efficiency in the context of complex PTKR.

Quantum-inspired algorithms (QIAs) are computationally efficient algorithms that generate a quantum model in classical hardware as an alternative to the use of quantum models. In this research paper, the researchers carried out an experimental appraisal of four quantum-inspired algorithms in quantum-inspired Evolutionary Algorithm (QEA), quantum-behaved Particle Swarm Optimization (QPSO), quantum-inspired Genetic Algorithm (QIGA) and quantum Simulated Annealing (QSA) on benchmark optimization functions. The introduction of a simulation-based methodology was done in Python to test the convergence speed, global search capability, accuracy and robustness. According to the experimental findings, QPSO always performs well compared with other methods in the achievement of faster convergence and enhanced global optimality, but QEA is highly successful in cases of discrete combinatorial problems. The paper concludes that quantum-inspired models are effective in modeling the key quantum concepts like probabilistic superposition and tunneling to attain high quality optimization results with conventional computing platforms. Such results demonstrate the opportunities of QIAs as scalable pre-quantum tools of designing, AI, and industrial optimization.

With the rapid advancement of artificial intelligence (AI), reducing the fan-in-induced current and power overhead in synaptic arrays has become a key challenge for neuromorphic hardware. This work proposes a NOR flash-based double-gate tunneling field-effect transistor (TFET) synaptic cell that directly addresses this issue. By exploiting the ultralow read current of TFET, the proposed device mitigates excessive cumulative current under large fan-in while maintaining sufficient read margin. The double-gate structure provides independent access paths for the top and bottom gates, enabling cell-level potentiation and depression within NOR architecture. Using an Incremental Step Pulse Programming/Erasing scheme, the device achieves highly linear 256-level analog weight modulation, which is then mapped to a positive-weight neural network model through a dedicated weight-shift and quantization strategy. The practicality of this approach is validated by reproducing a convolutional neural network (CNN) for Fashion-MNIST image classification on the proposed synaptic hardware model, where the inference accuracy closely tracks that of the original software model and even improves by approximately 0.06%. The reproduced accuracy closely matches that of the original software model without noticeable degradation, demonstrating its potential applicability in real-world scenarios. These results demonstrate that the proposed TFET-based double-gate flash synapse provides a viable device and mapping framework for low-power, high-fan-in neuromorphic systems.

Non‐dispersive infrared gas sensors, renowned for their high selectivity and high reliability, are extensively employed in applications of smart agriculture. In particular, a stable and high‐emission infrared source component plays a critical role in the proper functioning of non‐dispersive infrared systems. However, current infrared sources usually have shortcomings in poor temperature homogeneity within the active area and low‐power consumption. Here, we demonstrate a wafer scale, in situ integrated infrared source combined with an Al@NF‐based radiation layer, achieving a high emissivity of 0.8 at 4.26 μm. Through iterative optimization of the microheater pattern, the temperature homogeneity reaches an impressive 90%. In the integrated Al@NF‐infrared source sensing system, the power density is reduced from 386.8 to 256.7 mW/mm 2 ; meanwhile, its operational efficiency is increased eighteenfold, from 0.39% to 7.24%. The developed device enables precise tracking of greenhouse gas concentrations under controlled greenhouse conditions. The findings pave the way for low‐power non‐dispersive infrared systems and provide a new hardware model for smart agriculture.

The means of modeling mobile robots are considered, which can be used to develop algorithms for moving a robot in logistics systems. One of the possible solutions to this problem is the use of hardware and software tools for modeling robotic devices that allow simulating and controlling mobile robots in a virtual environment, therefore the purpose of this work is to analyze existing tools for hardware and computer modeling of warehouse systems based on mobile robots and autonomous vehicles, as well as to create models that make it possible to conduct research on such systems. The possibilities of hardware and software modeling are considered for the purpose of studying algorithms and equipment for moving mobile robots. The means of moving mobile robots using various navigation devices and taking into account the trajectory of movement are developed. The mock-ups of a mobile robot with a manipulator and an autonomous vehicle, as well as a lift based on the Arduino hardware and software complex, are developed. An algorithm for moving a mobile robot along a trajectory determined by moving along a straight line for a specified distance and turning by a specified angle is developed. A study of the means of movement of mobile robots and autonomous vehicles was conducted using the library and tools for creating models of the CoppeliaSim platform. The possibilities of using the obtained results for remote laboratory and practical classes were considered. The results of the work were used to create computer models of mobile robots for use in laboratory and practical classes. Based on the given examples of movement, tasks were created for conducting practical and remote classes to study the movement of mobile robots based on the Arduino hardware-software complex and the CoppeliaSim platform.

Pinching-antenna systems (PASS) present a break-through among the flexible-antenna technologies, and distinguish themselves by facilitating large-scale antenna reconfiguration, line-of-sight creation, scalable implementation, and near-field benefits, thus bringing wireless communications from the “last mile” to the “last meter”. To illustrate the benefits of PASS in next-generation wireless networks, a comprehensive tutorial is presented in this paper. First, the fundamentals of PASS are discussed, including PASS signal models, hardware models, power radiation models, and pinching antenna (PA) activation methods. Building upon this, the information-theoretic capacity limits achieved by PASS are characterized, and several typical performance metrics of PASS-based communications are analyzed to demonstrate its superiority over conventional antenna technologies. Next, the pinching beamforming design is investigated. The corresponding power scaling law is first characterized for the single-waveguide single-user case. For the joint transmit and pinching design in the general multiple-waveguide case, 1) a pair of transmission strategies is proposed for PASS-based single-user communications to validate the superiority of PASS, namely sub-connected and fully connected structures in terms of the connections between the baseband radio frequency chains and waveguides; and 2) three practical protocols are proposed for facilitating PASS-based multi-user communications, namely waveguide switching, waveguide division, and waveguide multiplexing. A possible implementation of PASS in wideband communications is further highlighted. Moreover, the channel state information (CSI) acquisition in PASS is elaborated with a pair of promising solutions, based on pilot-based channel estimation and beam training, respectively. To overcome the high complexity and suboptimality inherent in conventional convex-optimization-based approaches, machine-learning-based methods for operating PASS are also explored, focusing on selected deep neural network architectures and training algorithms. Finally, several promising applications of PASS in next-generation wire-less networks are highlighted to motivate future works.

Accurate channel estimation is critical for realizing the performance gains of massive multiple-input multiple-output (MIMO) systems. Traditional approaches to channel estimation typically assume ideal receiver hardware and linear signal models. However, practical receivers suffer from impairments such as nonlinearities in the low-noise amplifiers and quantization errors, which invalidate standard model assumptions and degrade the estimation accuracy. In this work, we propose a nonlinear channel estimation framework that models the distortion function arising from hardware impairments using Gaussian process (GP) regression while leveraging the inherent sparsity of massive MIMO channels. First, we form a GP-based surrogate of the distortion function, employing pseudo-inputs to reduce the computational complexity. Then, we integrate the GPbased surrogate of the distortion function into newly developed enhanced sparse Bayesian learning (SBL) methods, enabling distortion-aware sparse channel estimation. Specifically, we propose two nonlinear SBL methods based on distinct optimization objectives, each offering a different trade-off between estimation accuracy and computational complexity. Numerical results demonstrate significant gains over the Bussgang linear minimum mean squared error estimator and linear SBL, particularly under strong distortion and at high signal-to-noise ratio.

&lt;p&gt;Profit-Driven Generative Artificial Intelligence for Social Good: A Capitalist Lie and a Danger for Our Social Fabric&lt;/p&gt;

In this paper, we explore a low area-time elliptic curve point multiplication (ECPM) hardware accelerator over Curve25519. Firstly, we analyze the correlation of the data hazards of LADDER (main operation), the number of pipeline stages in modular multiplication (MM), and the number of iterations of multiplier in MM, and then establish an equation among them. Based on this equation, a new iterative hardware model of modular multiply-accumulator (MMAC) with stage=10 using a 128-bit multiplier, and a new LADDER scheduling are proposed to support high-performance scenarios for Curve25519 ECPM. Secondly, based on the proposed method, we propose a new 255-bit iterative MMAC. Finally, we explore the trade-off between area and time, present an efficient hardware design to accelerate the Curve25519 ECPM. The proposed accelerator is implemented on the Xilinx XC7Z020 platform. The results show that the proposed accelerator costs 2,367 Slices and 36 DSP blocks, with a clock frequency of 298 MHz. When calculating one Curve25519 ECPM, compared to previous designs with the best area*time (AT) value, the AT value of this design is 220% superior to their design, while the area and time have improved by 52% and 108% compared to theirs, respectively.