Search Results (404)

Analog in-memory computing (AIMC) has emerged as a promising approach to mitigate the Von Neumann bottleneck in matrix operations, which are common in deep learning applications. However, the practical implementation of resistive crossbar arrays is limited by challenges in signed weight representation, conductance quantization, and device nonlinearity. This paper presents a differential mixed-signal architecture for accurate signed matrix–vector multiplication (MVM), integrated with a RISC-V microcontroller for edge inference applications. A structured digital-to-analog mapping framework encodes quantized neural network weights into programmable conductance values while preserving arithmetic correctness. The design employs voltage-mode input encoding, differential current summation, and transimpedance-based readout followed by analog-to-digital conversion, enabling single-cycle signed accumulation without duplicating crossbar resources. A 32 × 16 dual-layer prototype crossbar was fabricated and experimentally characterized. Measurements demonstrate a mean absolute percentage error (MAPE) below 1% within the linear operating region and below 4% over the full-scale conductance range. These results validate the robustness of the proposed mapping methodology and confirm the feasibility of hybrid analog–digital acceleration for edge AI systems. Consequently, this discrete prototype serves as a physical verification platform for the AIMC approach, providing valuable insights for more efficient mixed-signal computing integrated circuit (IC) designs. Full article

Small unmanned aerial vehicles are difficult short-range radar targets because their millimeter-wave radar cross-sections often fall between −10 and −25 dBsm. This paper presents a modular analytical and simulation-based benchmark of a cascaded 77 GHz TDM-MIMO FMCW radar with 12 transmitters and 16 receivers, yielding a 192-element virtual ULA over a 40 m instrumented range. The framework is organized around the main counter-UAV sensing functions: range–Doppler processing first evaluates target observability and provides range–Doppler gates; Doppler-dependent TDM phase compensation is then required before virtual-array snapshots are formed for DoA estimation; and a separate long-dwell single-transmitter branch evaluates micro-Doppler separability using handcrafted features and a nearest-centroid Mahalanobis classifier. Four benchmarks are considered: detection under Swerling fluctuation models, residual TDM phase error caused by Doppler quantization, DoA estimation under an idealized far-field snapshot model, and micro-Doppler separability among UAV and bird classes. Under Swerling I, targets with a mean RCS of $-10$ dBsm or larger maintain detection probability above 0.9 throughout the 40 m window, whereas the $-20$ and $-25$ dBsm classes fall below that level at about 28 m and 21 m. In the far-field DoA benchmark, TLS-ESPRIT gives the lowest conditional RMSE and remains about 13–14 dB above the subarray CRLB at moderate SNR; however, these angular results are reference ceilings because the short-range operating region violates the full-aperture far-field condition and because residual TDM phase error can be severe without accurate compensation. In the micro-Doppler benchmark, birds exceed 95% per-class accuracy at 20 dB total SNR, but overall four-class accuracy saturates near 72–75% and UAV-only three-class accuracy near 63%, with most confusion between the micro-quadrotor and fixed-wing classes. This study therefore identifies architecture-specific performance margins and limitations before measured-data field validation, rather than claiming complete deployment-level performance. Full article

Spaceborne neural processing units (NPUs) increasingly support real-time deep-learning inference, but their dense multiply-accumulate arrays are vulnerable to radiation-induced soft errors. Conventional radiation-hardening methods improve reliability through hardware redundancy, but they incur substantial area, performance and compiler-mapping overheads. This paper proposes tile-resilient algorithm-based fault tolerance (TR-ABFT), a software-scheduled, detection-oriented scheme for quantized NPU inference. TR-ABFT generates checksum information at tile granularity and maps checking tasks onto the original processing element (PE) array without changing the hardware topology. To make ABFT compatible with INT8 datapaths, we design two checksum-coding strategies: checksum decomposition and modulo-239 checksum coding. The modulo-239 scheme removes structural missed detections for two-bit flips with bit-position spacings in (1, 31), while preserving compatibility with signed INT8 inputs. Evaluations on ResNet, YOLOv8, and RT-DETR show that, on a $16\times 16$ array, TR-ABFT introduces only 6.37% to 24.61% additional computational overhead. By converting spatial redundancy into schedulable temporal redundancy, TR-ABFT preserves systolic-array regularity and provides a low-overhead reliability-enhancement mechanism for space-grade neural-network accelerators. Full article

This paper investigates the output synchronization of fractional-order cyber-physical systems (FOCPSs) under communication constraints. To address limited bandwidth and high transmission costs, an event-triggered encoding-decoding sampled-data iterative learning control (ET-EDSDILC) protocol is proposed. The control law integrates a quantized sampling framework with an encoding–decoding mechanism to reconstruct control signals and address communication constraints. Furthermore, an event-triggered mechanism based on error energy attenuation (EEA) is developed to adjust communication frequency by monitoring error trends, thereby reducing unnecessary data transmissions. By applying fractional-order calculus and the contraction mapping principle, sufficient conditions for output synchronization are derived. Numerical simulations show that the proposed ET-EDSDILC framework reduces communication overhead and data redundancy while maintaining tracking performance, offering a solution for FOCPSs under communication constraints. Full article

Despite recent advances, medical question answering systems still struggle with domain-specific reasoning and data efficiency. This paper presents Med-LLaMA3, a family of medical large language models developed by parameter-efficient fine-tuning of the LLaMA-3.1 (8 billion) and LLaMA-3.2 (1 and 3 billion) architectures using quantized low-rank adaptation (QLoRA) and low-rank adaptation (LoRA) with 4-bit quantization. Beyond model training, this work contributes the following: (1) a formalized dataset curation taxonomy (source type × clinical granularity × task format) with a source-category ablation confirming that the multi-source combination drives benchmark gains beyond any single category; (2) a systematic characterization of low-rank-adaptation rank-scaling behavior for the LLaMA-3 family in the medical domain (monotonic improvement up to rank 128, with no observed plateau); and (3) statistically validated comparisons using McNemar’s test and 95% bootstrap confidence intervals. We curated a medical instruction dataset of over 1.5 million samples spanning medical examinations, clinical dialogues, and biomedical literature. Our approach trains only ∼4% of the base model’s parameters and, consistent with prior studies of parameter-efficient methods in the medical domain, achieves performance comparable to full fine-tuning at a fraction of the memory footprint. Evaluated with five in-context examples per prompt, the 8-billion-parameter model attains a mean accuracy of 75.71% across the eight medical-domain subsets of the Massive Multitask Language Understanding benchmark; improvements over the unmodified LLaMA-3.1-8B-Instruct baseline are statistically significant on the medical multiple-choice benchmark MedMCQA and, after Bonferroni correction across the eight subsets, on three subsets (Clinical Knowledge, Medical Genetics, and Nutrition), with two further subsets being significant only before correction. A structured named-entity-recognition evaluation on 100 hospital discharge summaries (macro-averaged F1 0.94; dual-annotator agreement $\kappa =0.87$) provides complementary evidence of clinical-text utility. A safety mitigation pilot shows that context-disambiguation preprocessing reduces the highest-severity abbreviation-ambiguity error rate from 30% to 10% on a 30-case held-out set. These results show that parameter-efficient fine-tuning can deliver high-performance medical large language models while training only ∼4% of the model’s parameters and reducing memory use by roughly 75%, enabling development on low-cost consumer-grade hardware. Full article

Blood typing remains a manual, subjective procedure when not reliant on centralized laboratory infrastructure. This study presents an automated blood typing system for point-of-care deployment, developed in collaboration with CRIAM, whose portable device captures reaction images for in vitro diagnostics. The system integrates computer vision and artificial intelligence to classify these reactions automatically. Fourteen classification pipelines were trained and evaluated with a 3090-image dataset, encompassing fine-tuned convolutional neural networks, raw pixel-based classifiers, and hybrid architectures pairing pretrained embeddings from DINOv2 and EfficientNet-B4 with lightweight classifiers. Embedding-based approaches consistently outperformed alternatives in accuracy and cross-fold stability. The best pipeline, in terms of performance and suitability for low-power devices, combined DINOv2-small embeddings with logistic regression, achieving 99.87 ± 0.12% mean accuracy. After 8-bit integers (INT8) quantization and retraining with data augmentation, accuracy improved to 99.97 ± 0.03%, surpassing the uncompressed baseline. All misclassifications were traced to borderline weak-positive Rh/D reactions, confirming errors are localized and explainable. Held-out validation on 856 images yielded 99.53% accuracy, with the single error attributed to a lighting artifact. While deployment on a legacy 32-bit CPU prototype processes four images in approximately 4.7 min, hardware benchmarking confirmed feasibility, from a Raspberry Pi Zero 2W to high-end mobile processors. These results establish quantized embedding-driven architectures as a solution for automated blood typing in point-of-care and resource-limited settings. Full article

Future wireless systems such as 5G-Advanced and 6G are expected to rely increasingly on multi-user MIMO and distributed multi-antenna transmission, where accurate channel direction information (CDI) is essential for interference management. In limited-feedback downlink systems, however, finite-rate CDI feedback introduces quantization error, resulting in residual interference and rate loss in zero-forcing beamforming. This paper proposes a quantization-error-threshold-based user admission scheme for limited-feedback MU-MIMO downlink systems. In the proposed scheme, each user feeds back its quantized CDI and channel quality information only when its CDI quantization error is below a predefined threshold, and the base station performs semi-orthogonal user selection and zero-forcing beamforming over the admitted users. The proposed threshold controls the tradeoff between feedback-overhead reduction and candidate-user availability while improving the reliability of the CDI used for precoding. An analytical framework is developed to characterize the threshold-dependent scheduled-user count, ergodic sum-rate, and feedback overhead. Simulation results show that the proposed scheme improves the sum-rate compared with conventional SUS and substantially reduces the feedback overhead, especially as the number of users increases. Full article

In this review paper, we present a framework for the characterization of optimal decision rules in M-ary hypothesis-testing problems where the performance metric is defined as a function of pairwise error probabilities. This framework is based on the approaches developed in several recent studies in the literature, which are unified and presented in a tutorial fashion in this paper. A pairwise error probability represents the probability of selecting a specific hypothesis when a different hypothesis is true, and can be stacked into a pairwise probability vector for a given problem. In the considered framework, instead of optimizing the performance metric of interest over the infinite-dimensional set of all possible decision rules, the optimization is performed directly over the compact and convex set of all achievable pairwise probability vectors. We demonstrate that any pairwise probability vector within this feasible set can be realized via a randomization of at most two likelihood ratio quantizers (LRQs) with different sets of parameters. While one of these LRQs can always be selected as a deterministic LRQ, the other one is possibly a randomized LRQ, which can be written as a randomization of at most $M(M-1)$ deterministic LRQs, with M denoting the number of hypotheses. The main advantage of this framework is that it allows for the attainment of pairwise probability vectors that do not reside on the boundary of the feasible set and that are fundamentally inaccessible via LRQs, which are optimal for classical performance metrics such as the Bayes risk or the Neyman–Pearson criterion. Furthermore, we show that the characterization of decision rules with the presented framework is particularly advantageous for performance metrics based on prospect theory (PT), such as behavioral utility. Specifically, it is demonstrated that the optimal pairwise probability vector for a PT-based metric is not guaranteed to lie on the boundary of the feasible set of pairwise probability vectors. This results in suboptimal performance achieved by LRQs for such performance metrics. On the other hand, the randomized decision rules characterized in this paper can achieve pairwise probability vectors located in the interior of the feasible set, thereby yielding optimal performance. Numerical results corroborate these findings, demonstrating that the decision rules characterized within our framework yield optimal behavioral utility-based performance scores. Full article

This paper investigates the problem of non-fragile dissipative filtering for discrete-time nonlinear networked systems with dynamic quantization, a dynamic event-triggered mechanism and stochastic cyber attacks. The nonlinear networked system under investigation is described by an uncertain Takagi–Sugeno (T-S) fuzzy model. In this work, a novel fuzzy-dependent dynamic event-triggered communication scheme and the dynamic quantization strategy, integrated with an online adjustment rule, are introduced to reduce the frequency and volume of data transmission, thus realizing more rational utilization of the limited communication resources. In addition, the stochastic cyber attacks are characterized by a random variable obeying the Bernoulli distribution. The core focus of this paper is to design a non-fragile filter such that the resulting filtering error system is stochastically stable and meets the prescribed dissipative filtering performance. Based on the matrix inequality decoupling technique, the design conditions of the desired filter are derived and presented in the form of linear matrix inequalities (LMIs). Finally, the effectiveness and superiority of the proposed filter design approach is verified via two simulation examples. Full article

Real-time deployment of modern object-detection networks on edge devices is challenging because of limited compute resources, external-memory bandwidth, and strict power constraints. To address these issues, this paper presents a host–FPGA collaborative accelerator for quantized YOLOv5n on a Xilinx Zynq-7100 platform. The proposed design includes a modular multi-operator neural processing unit supporting seven atomic operators, a Dual-Multiply Packing (DMP) scheme to improve DSP48E1-based INT8 convolution density, a cache–compute–cache dataflow with global ping-pong buffering to overlap DMA transfers and computation, and a Multi-Quantization Domain Alignment (MQDA) pipeline to preserve accuracy at Add and Cat fusion nodes. Implemented at 200 MHz, the prototype achieves 24.617 ms FPGA-side forward-inference latency, 36.686 ms end-to-end single-frame latency, 27.2 FPS system-level performance, 182.8 GOPS equivalent throughput, and 8.536 W on-chip power consumption, corresponding to 21.42 GOPS/W. Experimental results also show that INT8 quantization causes only limited accuracy degradation, while MQDA improves quantized detection accuracy by reducing cross-domain fusion error. These results demonstrate that the proposed architecture provides an effective balance among throughput, energy efficiency, hardware cost, and quantized accuracy for real-time edge object detection. Full article

This paper presents a passive residue-coupled discrete-time delta–sigma ($\Delta \Sigma$) modulator for low-power narrowband sensing applications. Instead of adding a fourth active integrator, the proposed architecture keeps a third-order switched-capacitor main loop and reuses the intrinsic top-plate residue of an 8-bit asynchronous successive-approximation-register (SAR) quantizer. The retained capacitive digital-to-analog converter (CDAC) residue is passively reinjected through a charge-redistribution path, introducing an additional high-pass error-propagation factor in the effective noise transfer function (NTF). Under a bounded effective coupling coefficient, the proposed loop approaches fourth-order-like in-band noise suppression while retaining third-order active-loop complexity. Behavioral simulations show that the Enhanced mode improves the peak signal-to-noise-and-distortion ratio (SNDR) by 16.9 dB over the Baseline third-order mode at an oversampling ratio (OSR) of 128. Circuit-level corner verification of the standalone SAR confirms correct bit cycling and a settled residue-retention window under typical–typical (TT), slow–slow (SS), and fast–fast (FF) conditions: with the slowest conversion window of about 21.4 ns at the SS corner and a sampling period of 39.06 ns at $f_s=25.6$ MHz, roughly 17.66 ns of timing margin remains for residue holding, passive reinjection, and clock non-overlap. The proposed method provides an architecture-level route for improving in-band noise shaping without increasing the number of active integrator stages, and is particularly attractive for low-power, narrowband, and sensor-oriented analog-to-digital converter (ADC) applications. Full article

Background and Objectives: Dental procedure coding is essential for accurate billing, reimbursement, and clinical documentation, yet it remains largely manual, time-consuming, and error-prone. While natural language processing (NLP) has enabled significant advances in automated medical coding, limited work has focused on the dental domain, particularly the assignment of Code on Dental Procedures and Nomenclature (CDT) codes from free-text clinical notes. This study aims to develop and evaluate an artificial intelligence framework that integrates large language models (LLMs) and traditional deep learning methods to automate CDT code extraction from narrative dental documentation. Methods: We evaluated three LLM-based strategies—zero-shot prompting, QLoRA fine-tuning, and parameter-efficient fine-tuning (PEFT) using LoRA—alongside a supervised Bidirectional GRU (Bi-GRU) classifier. Experiments were conducted using a synthetic dataset designed to emulate real-world dental encounters. Structured JSON output schemas, few-shot prompting, and scalable batch inference pipelines were employed to ensure consistent and interpretable predictions. Model performance was assessed using micro- and macro-averaged F1 scores, precision, recall, exact-match accuracy, and Hamming loss. Results: The zero-shot LLM achieved the highest micro-F1 score (0.9614) and perfect recall for frequent CDT codes, demonstrating strong baseline reasoning without task-specific training; however, performance declined for rare procedures and diagnostic code hallucinations were common. Fine-tuning improved domain alignment, with the non-quantized PEFT LoRA model outperforming QLoRA across all metrics, though both fine-tuned LLMs showed tendencies to over-generate plausible but incorrect codes. The Bi-GRU model achieved balanced performance (micro-F1 = 0.9362, macro-F1 = 0.9377) with minimal hallucinations but occasionally missed context-dependent procedures. Conclusions: These findings highlight complementary strengths between LLM-based and supervised approaches. LLMs provide strong contextual understanding and rapid deployment, while traditional models offer stable and precise multi-label classification. This work supports the development of hybrid, schema-constrained systems for scalable dental procedure coding. Full article

Background: Spiking neural networks (SNNs) have attracted significant attention in the field of brain–computer interfaces owing to their distinctive biological plausibility and energy efficiency advantages. However, the discrete nature of spikes renders gradient-based differentiation infeasible, making it difficult to directly obtain well-trained SNNs. A common approach is to transfer the weights from artificial neural networks (ANNs) to SNNs. However, this process introduces conversion errors that pose significant challenges. Methods: To address these challenges, we propose the self-rectifying integrate-and-fire (SRIF) neuron, which employs negative spikes to reduce asynchronism error and rectification spikes to diminish clipping error. Concomitantly, we propose a collaborative trim (CT) training framework that introduces a quantized network to perceive the weights and results of SNNs, which can further improve performance. Result: The proposed training methodology enables SNNs to achieve performance metrics comparable to those of ANNs in EEG-based motor imagery (MI) classification. Conclusions: Experimental results demonstrate that our method not only preserves the superior classification performance of ANNs but also leverages the superior energy efficiency and lower computational complexity of SNNs. Full article

Ion-acoustic waves in dense quantum plasmas are strongly influenced by Fermi degeneracy, Landau quantization, and finite-temperature effects, and in many relevant environments, they also experience memory and nonlocal transport processes that cannot be captured within the planar integer Korteweg-de Vries (KdV) paradigm. In the present work, we revisit this problem by considering a two-fluid, partially degenerate electron-ion plasma in which electron trapping in the presence of a quantizing field and finite temperature is taken into account. Starting from the normalized fluid-Poisson system appropriate for such magnetized quantum plasmas, the reductive perturbation technique is used to derive the planar integer KdV equation for weakly nonlinear ion-acoustic disturbances. Within this integer-order KdV framework, we recast the evolution equation as a planar dynamical system, construct the associated Hamiltonian and effective Sagdeev-like potential, and demonstrate the existence of compressive solitary waves and nonlinear periodic modes via homoclinic and periodic phase-space orbits. Exact multi-soliton solutions and interaction states are then obtained by combining Hirota’s direct bilinear method with generalized Wronskian representations, allowing us to describe not only standard one-, two-, and three-soliton profiles but also positon-negaton interactions relevant to magnetized, partially degenerate plasmas. To incorporate hereditary and history-dependent effects that arise from anomalous transport and nonlocal temporal response in dense environments, we extend the model by introducing a Caputo time-fractional derivative, thereby obtaining a time-fractional KdV (FKdV) equation that continuously connects the classical KdV limit to fractional dynamics. The FKdV equation is analyzed using the Tantawy technique. This semi-analytical iterative scheme yields rapidly convergent series approximations for the fractional ion-acoustic soliton and provides explicit control of the approximation error. The fractional solutions show that varying the order of the Caputo derivative modifies the amplitude, width, and temporal relaxation of the solitary structures and can even split the pulse into two distinct lobes, in contrast with the nearly rigid propagation predicted by the integer-order KdV equation. Taken together, these results clarify how Landau quantization, finite electron temperature, and fractional-order memory jointly shape the morphology, robustness, and interaction properties of ion-acoustic structures in strongly magnetized quantum plasmas of astrophysical and high-energy-density laboratory interest. Full article

Sensorless permanent magnet synchronous motor (PMSM) drives rely on accurate rotor electrical angle and speed estimation, vulnerable to noisy currents, quantization, and sensor biases. Fixed-bandwidth phase-locked loops (PLLs) entail an intrinsic trade-off between fast transient tracking and high-frequency noise rejection. This paper proposes an adaptive PLL based on linear active disturbance rejection control (LADRC), where a virtual coordinate formulation treats electrical-angle mismatch as a lumped disturbance estimated online by a linear extended state observer (LESO). The observer bandwidth dynamically adapts to the LESO innovation. To optimize performance, adaptive-law parameters are tuned offline via success-history adaptive differential evolution with linear population size reduction (L-SHADE). Comparative simulations against a proportional-integral PLL indicate substantially improved robustness to measurement noise, analog-to-digital quantization, and current-sensor DC offset. Specifically, the speed root-mean-square error decreases from $68.9\mathrm{r}/\min$ to $20.7\mathrm{r}/\min$ under $0.15\mathrm{A}$ additive noise, and from $1.55\mathrm{r}/\min$ to $0.48\mathrm{r}/\min$ under 12-bit quantization at $200\mathrm{r}/\min$. These enhancements reduce reliance on high-precision sensing hardware, offering a practical solution for low-cost, highly reliable motor control in complex industrial environments. Full article