This article presents a method of sickle cell detection from microscopic images. We extract five attribute values from the connected components of an image, and train machine learning classifiers to recognize the sickle cells. Four classifiers were experimented with and the vest one was the K-Nearest neighbor classifier with 97.3% accuracy. The other classifiers are the Neutral network, Decision tree and Naïve Bayesian classifiers which resulted in accuracy rates of between 89-96.3%. This method is applicable for use in low cost computers since it is computationally cheap. The findings of this research can be considered as a screening method for diagnosing sickle cell aneamia.

The demand for edge machine learning (ML) in Internet of Things (IoT) applications has driven interest in Microcontroller Unit (MCU)-based TinyML solutions, especially with the rise of the RISC-V Instruction Set Architecture. While MCUs are power-efficient, their limited resources challenge the deployment of complex ML models. Mixed-Precision Quantization (MPQ) can achieve the best trade-off between model size, energy consumption and accuracy, by using different precision across model layers. However, MCU-class processors often lack hardware support for MPQ. We present an end-to-end flow from training to hardware deployment designed to efficiently run Mixed-Precision Quantized Neural Networks (MP-QNNs) on MCU-class processors. Central to our approach is STAR-MAC, a precision-scalable Multiply-and-Accumulate unit that supports flexible MPQ operations on 16-, 8-, and 4-bit integer data. STAR-MAC combines two subword-parallel techniques, Sum-Together and Sum-Apart, in a unified multiplier architecture that effectively reconfigures for Fully-Connected, 2D Convolution, and Depth-wise layers. We integrate STAR-MAC into the low-power RISC-V Ibex core and validate our flow using an FPGA-based System-on-Chip setup. Inference results on MLPerf Tiny MP-QNN models using our modified TensorFlow Lite for Microcontrollers (TFLM) deployment flow show a 68% latency reduction with little to no accuracy drop against their 8-bit counterparts using the standard TFLM runtime. Synthesis on a 28-nm CMOS technology indicates limited area and power overhead over the original Ibex. We open-source our framework to foster MP-QNN deployment on MCU-class RISC-V processors for low-power and low-latency IoT data processing.

Mixed-precision quantization (MPQ) is gaining momentum in academia and industry as a way to improve the trade-off between accuracy and latency of Deep Neural Networks (DNNs) in edge applications. MPQ requires dedicated hardware to support different bit-widths. One approach uses Precision-Scalable MAC units (PSMACs) based on multipliers operating in Sum-Together (ST) mode. These can be configured to compute N = 1, 2, 4 multiplications/dot-products in parallel with operands at 16/N bits. We contribute to the State of the Art (SoA) in three directions: we compare for the first time the SoA ST multipliers architectures in performance, power and area; compared to previous work, we contribute to the portfolio of STbased accelerators proposing three designs for the most common DNN algorithms: 2D-Convolution, Depth-wise Convolution and Fully-Connected; we show how these accelerators can be obtained with a High-Level Synthesis (HLS) flow. In particular, we perform a design-space exploration (DSE) in area, latency, power, varying many knobs, including PSMAC units parallelism, clock frequency and ST multipliers type. From the DSE on a 28-nm technology we observe that both at multiplier level and at accelerator level there is no one-fits-all solution for each possible scenario. Our findings allow accelerators' designers to choose, out of a rich variety, the best combination of ST multiplier and HLS knobs depending on the target, either high performance, low area, or low power.

In Deep Neural Networks (DNN), the depth-wise separable convolution has often replaced the standard 2D convolution having much fewer parameters and operations. Another common technique to squeeze DNNs is heterogeneous quantization, which uses a different bitwidth for each layer. In this context we propose for the first time a novel Reconfigurable Depth-wise convolution Module (RDM), which uses multipliers that can be reconfigured to support 1, 2 or 4 operations at the same time at increasingly lower precision of the operands. We leveraged High Level Synthesis to produce five RDM variants with different channels parallelism to cover a wide range of DNNs. The comparisons with a nonconfigurable Standard Depth-wise convolution module (SDM) on a CMOS FDSOI 28-nm technology show a significant latency reduction for a given silicon area for the low-precision configurations.

The presence of foreign bodies in packaged food is a serious concern for both final consumers (allergies, injuries, choking) and food manufacturers (reputation and economic losses). In particular, low-density plastics, glass and wood splinters are hard to detect even by the most advanced X-ray imagers. One solution is Machine-Learning-based Microwave Sensing (MLMWS): a non-invasive, contactless, and real-time method which uses a machine-learning (ML) classifier to analyze the scattered microwaves from the irradiated target object. In this paper, we want to extend our previous work about contaminant detection in cocoa-hazelnut spread jars by proposing an enhanced ML flow to increase the accuracy of the ML classifier. For the first time in this case study, we use a multi-class classifier, we train it with scattering parameters measured at multiple microwave frequencies, with a new pre-processing scaler, data augmentation, quantization-aware training and a pruning schedule. The results show a contaminant detection multi-class accuracy of 94.167% with a latency of 26 µs when targeting an AMD/Xilinx Kria K26 FPGA. Finally, we released our datasets publicly to OpenML.1

This study proposes and evaluates a novel implantable antenna testing method using a tissue gel phantom combined with a living human subject to simulate realistic implantation conditions. The gel phantom, shaped by 3D-printed molds and composed of water, sugar, salt, and agar, enables antenna performance analysis through channel gain and Received Signal Strength Indicator (RSSI) measurements in mobile environments. By incorporating human subjects, we capture channel gain variations in dynamic conditions influenced by design, gain, implantation site, and subject movement. This novel setup allowing rotational movement and untethered Bluetooth Low Energy (BLE) packet analysis reveals a previously undetectable 20 dB difference between line-of-sight and non-line-of-sight scenarios, highlighting the limitations of traditional bench-top phantoms.

To emulate a quantum computation on a classical computer i.e. the evolution of the unitary operations on the wave function of the particle in quantum mechanics, we have to perform unitary matrix and normalized vector multiplications in the high-level programming languages of Python, C++, Java, etc. Quantum Libraries already available perform the matrix-vector multiplication in the backend using the numpy libraries of Python like Qiskit or use a C++ wrapper to further optimize the runtime it as in Qiskit-Aer Simulators. Since a fully functioning fault-tolerant computer is decades away, it is in our best interest to design new quantum algorithms and develop accelerator test beds for Quantum Emulations. All the quantum computer operations can be emulated on a classical computer, with the only downside being that the matrix multiplications scale up as 𝑂 (𝑁 3) in runtime. In contrast, the quantum computer scales it up as 𝑂 (𝑁 2 𝑙𝑜𝑔 2 𝑁), where N = 2 𝑛 , where N is the matrix dimension, where n is the number of qubits, so the runtime for quantum emulations on the classical computer increases exponentially with increase in number of qubits and increases linearly with increase in number of depths, complexity wise. Though it is not possible to change the exponential index, it is possible to reduce the runtime for quantum emulations on classical computers by use of GPU and Alveo Accelerator Cards, and also code optimization on the software side like using a C++ wrapper. In this paper, we will benchmark the matrix-matrix multiplications on HPC Accelerator Cards varying the qubit size and the depth of the quantum circuit and provide a universal mathematical equation for the runtime on the GPU and Alveo Vector Cards for two variables of qubit size and quantum circuit depth. So a theoretical limit on qubit size and qubit depth exactly can be established for quantum emulations on present classical supercomputers.

In this letter, an error is identified in Proposition 3 presented in Björnson et al. (2020) which includes an erroneous expression to compute the optimal number of reflecting elements that minimizes the total power consumption and maximizes the energy efficiency in intelligent reflecting surface (IRS)-supported transmission. The correct version of this expression and a few comments are provided in this letter, and simulation results are also presented to support this claim.

As digital infrastructures continue to expand globally, organizations and individuals alike face an ever-growing threat from ransomware attacks, which disrupt critical systems and hold sensitive data hostage. The development of an adaptable and precise detection mechanism is crucial to counter the sophisticated and evolving tactics employed by ransomware actors. This research introduces the Dynamic Anomaly Matrix Embedding (DAME) framework, a highly novel approach designed to detect ransomware through a combination of matrix embedding techniques and dynamic behavioral profiling, achieving a substantial advancement over traditional detection models. DAME’s architecture allows for the real-time monitoring and analysis of system behaviors, effectively differentiating ransomware activity from legitimate processes with minimal false positives and negatives. Evaluation results demonstrate that DAME not only maintains a high detection accuracy and precision but also optimally balances processing speed and resource utilization, proving effective even in resource-constrained environments. Comparative analysis further reveals that DAME surpasses traditional detection models, such as Support Vector Machines and Random Forest, in both efficiency and adaptability to novel ransomware variants. The presented findings affirm DAME’s robustness as a cybersecurity solution and underscore its significant contributions to advancing autonomous ransomware detection methodologies.

The integration of low Earth orbit (LEO) satellite communication into 6G networks promises a transformative impact on global connectivity by expanding coverage to remote regions and enhancing service reliability. However, this new infrastructure also introduces significant security challenges due to its expansive attack surface. To address this concern, we propose a dynamic security functions allocation (SFA) model that optimizes the allocation of security functions (SFs) across satellites while considering computational resource limitations, dynamic topology changes, and the visibility constraints of satellite constellations. Our model leverages the flexibility of 6G network slicing (NS) to share non-critical SFs between slices, reducing resource overhead while maintaining essential security demands. To minimize the risk of sharing highly sensitive SFs between slices, our model employs a nonlinear penalty, which prioritizes minimizing risk by aggressively penalizing high-risk SFs sharing. This dynamic risk management framework assesses the probability and impact of security breaches, ensuring that SFs are shared only when the security risk is acceptable, balancing resource efficiency and security. By dynamically adapting to the network's operational conditions, our approach provides a robust framework for efficient and secure satellite communication in 6G networks. Simulation results demonstrate the model's flexibility in managing trade-offs across key network performance metrics.

Sensing-assisted communication is critical to enhance the system efficiency in integrated sensing and communication (ISAC) systems. However, most existing literature focuses on large-scale channel sensing, without considering the impacts of small-scale channel aging. In this paper, we investigate a dual-scale channel estimation framework for sensing-assisted communication, where both large-scale channel sensing and small-scale channel aging are considered. By modeling the channel aging effect with block fading and incorporating CRB (Cramér-Rao bound)-based sensing errors, we optimize both the time duration of large-scale detection and the frequency of small-scale update within each subframe to maximize the achievable rate while satisfying sensing requirements. Since the formulated optimization problem is non-convex, we propose a two-dimensional search-based optimization algorithm to obtain the optimal solution. Simulation results demonstrate the superiority of our proposed optimal design over three counterparts.

Escalating threats from increasingly sophisticated ransomware attacks pose significant risks to data integrity, financial stability, and operational continuity for organizations across sectors. Addressing these threats requires a novel detection framework that overcomes the limitations of traditional signature-based and machine learning-based approaches, particularly their vulnerability to adaptive ransomware tactics and resource constraints. The proposed Dynamic Threat Imprint Analysis (DTIA) method leverages real-time behavioral analysis combined with adaptive learning mechanisms, enabling it to detect both known and previously unseen ransomware variants with high accuracy. Experimental evaluation demonstrates that DTIA achieves a 98.7% detection rate for known ransomware while effectively adapting to new variants, maintaining a 95% detection rate for previously unseen threats. Additionally, DTIA operates with minimal impact on system performance, showcasing scalable and efficient detection capabilities suited for deployment in environments with diverse computational resources. These findings demonstrate the potential of DTIA as a sophisticated and adaptable solution for the evolving challenge of ransomware detection, enhancing cybersecurity resilience through advanced behavioral analysis and dynamic learning capabilities.

In the context of high fossil fuel consumption and inefficiency within China's energy systems, effective demand-side management is essential. This study examines the thermal characteristics of various building types across different functional areas, utilizing the concept of body coefficient to integrate their unique structural and energy use traits into a demand response framework supported by real-time pricing. We developed a Stackelberg game-based bi-level optimization model that captures the dynamic interplay of costs and benefits between integrated energy providers and users. This model is formulated into a Mixed Integer Linear Programming (MILP) problem using Karush-Kuhn-Tucker (KKT) conditions and linearized with the Big M method, subsequently solved using MATLAB and CPLEX. This approach enables distinctive management of heating loads in public and residential areas, optimizing energy efficiency while balancing the interests of both providers and users. Furthermore, the study explores how the proportion of different area types affects the potential for reducing heat loads, providing insights into the scalability and effectiveness of demand response strategies in integrated energy systems. This analysis not only highlights the economic benefits of such strategies but also their potential in reducing dependency on traditional energy sources, thus contributing to more sustainable energy system practices.

Recent research in the field of Large Language Models (LLMs) has given a new direction to the capabilities of AI agents for solving complex problems. This paper attempts to explore one such use case to investigate LLMs-based AI agents' role in immersive technology, specifically focusing on GPT-4's vision capabilities in Augmented Reality (AR). The paper utilizes Smart App Agent Framework for recommending products. This recommendation system assists users to make context-aware decisions during their online shopping experience.

This paper introduces an advanced application of Large Language Models (LLMs) to predict Means of Compliance (MoCs) in aerospace defense systems, solely based on textual descriptions of system requirements. Amid increasing complexity and escalating demands on compliance verification processes, this study leverages a meticulously curated dataset of labeled requirements to train a fine-tuned model that automates MoC assignments. By incorporating machine learning classification techniques, the model demonstrates significant potential to enhance the efficiency and reliability of system verifications, markedly reducing the time and human effort traditionally required. A comparative analysis, incorporating feedback from Brazilian Aerospace Defense specialists, underscores the model's capability to match, and at times surpass, human accuracy in MoC identification, thereby supporting the development of more robust defense systems. This work not only contributes to the ongoing discourse on the integration of AI in systems development but also proposes a scalable solution to streamline compliance processes in the aerospace industry.

As ransomware attacks grow in complexity, traditional detection methods face significant challenges in identifying the advanced tactics, techniques, and procedures that define contemporary ransomware behaviors. Developing detection methodologies that capture the intricate and dynamic interactions of ransomware across encrypted networks presents an opportunity for innovation in cybersecurity. Hypergraph-Driven Anomalous Behavior Profiling (HDABP) offers a novel and scalable solution through representing complex network relationships in hypergraph structures, allowing for more precise detection of multi-faceted ransomware behaviors that evade conventional graph-based and signature-driven techniques. HDABP extends beyond simple pairwise relations, modeling high-order dependencies and unique interaction patterns within ransomwareinfected networks, thus revealing subtle anomalies associated with ransomware propagation and lateral movement. A comprehensive evaluation demonstrates HDABP's effectiveness through its high detection accuracy, resource efficiency, and adaptability in encrypted environments, where traditional detection methods are limited. Tested across multiple ransomware families and network conditions, HDABP achieves consistent low-latency performance and low false-positive rates, providing reliable detection capabilities crucial for real-time cybersecurity applications. HDABP's design also enables deployment within sectors that prioritize data privacy, such as finance and healthcare, while maintaining efficacy in recognizing evolving ransomware tactics without direct payload inspection. Through bridging theoretical hypergraph analysis with practical cybersecurity applications, HDABP represents a substantial advancement in ransomware detection, enhancing resilience against increasingly sophisticated ransomware threats.

Virtual Reality (VR) is increasingly used in neuropsychological assessments due to its ability to simulate real-world environments. This study aimed to develop and evaluate the Trail Making Test in VR (TMT-VR) and investigate the effects of different interaction modes and gaming skills on cognitive performance. A total of 71 young female and male adults (aged 18–35) with high and low gaming skills participated in this study. Participants completed the TMT-VR using three interaction modes as follows: eye-tracking, head movement, and controller. Performance metrics included task completion time and accuracy. User experience, usability, and acceptability of TMT-VR were also examined. Results showed that both eye tracking and head movement modes significantly outperformed the controller in terms of task completion time and accuracy. No significant differences were found between eye tracking and head movement modes. Gaming skills did not significantly influence task performance using any interaction mode. The TMT-VR demonstrates high usability, acceptability, and user experience among participants. The findings suggest that VR-based assessments can effectively measure cognitive performance without being influenced by prior gaming skills, indicating potential applicability for diverse populations.  

This scoping review examines the broad applications, risks, and ethical challenges associated with Extended Reality (XR) technologies, including Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR), within the context of Metaverse. XR is revolutionizing fields such as immersive learning in education, medical and professional training, neuropsychological assessment, therapeutic interventions, arts, entertainment, retail, e-commerce, remote work, sports, architecture, urban planning, and cultural heritage preservation. The integration of multimodal technologies-haptics, eye-tracking, face-and body-tracking, and braincomputer interfaces-enhances user engagement and interactivity, playing a key role in shaping the immersive experiences in the Metaverse. However, XR's expansion raises serious concerns, including data privacy risks, cybersecurity vulnerabilities, cybersickness, addiction, dissociation, harassment, bullying, and misinformation. These psychological, social, and security challenges are further complicated by intense advertising, manipulation of public opinion, and social inequality, which could disproportionately affect vulnerable individuals and social groups. This review emphasizes the urgent need for robust ethical frameworks and regulatory guidelines to address these risks while promoting equitable access, privacy, autonomy, and mental wellbeing. As XR technologies increasingly integrate with artificial intelligence, responsible governance is essential to ensure the safe and beneficial development of the Metaverse and the broader application of XR in enhancing human development.