Charge pump-based (CPB) dynamic biasing can be exploited in latched comparators to optimize the speed-noise tradeoff allowing operation at low supply voltage with high speed and low noise. In this work we provide the theoretical foundation for this optimization, and present two single-stage comparator topologies operating at 0.6 V supply. The former exploits dynamic biasing to achieve a delay of 95 ps with low power consumption and low noise, whereas the latter exploits a pseudo-differential approach to allow offset calibration with minimal overhead, at the cost of higher noise. Simulations in 28-nm CMOS validate the proposed model and show that the proposed topologies are able to achieve state-of-the-art performance.

It is envisioned that 6G mobile networks will enhance and majorly empower the Industry 4.0 paradigm, evolving towards smart factories with optimized and customized services. Especially the smart factory scenario with high-capacity data communication, which requires the usage of new portions of the electromagnetic spectrum (mmWave/sub-THz), is presenting us with new challenges, both in communications and networking. This article discusses the new challenges arising from highcapacity data communications in smart factories. It proposes extensions to the current Open Radio Access Network (O-RAN) standards for 6G networks to enable further evolution of Industry 4.0 and beyond. We motivate the need for real-time functionalities in O-RAN and an extended interface to the user equipment (UE) to allow for its fine-granular control.

Respiratory disorders have become the third largest cause of death worldwide, which can be assessed by one of the two key diagnostic modalities: breathing patterns (BPs) or the airflow signals, and respiratory sounds (RSs). In recent years, a few research has been conducted on finding correlation between these two modalities which indicate the structural flaws of lungs under disease condition. In this letter, we propose 'RS-2-BP': a unified deep learning framework for deriving the electrical impedance tomography-based airflow signals from respiratory sounds using a hybrid neural network architecture, namely ReSTL, that comprises cascaded standard and residual shrinkage convolution blocks, followed by feature refined transformer encoders and long-short term memory (LSTM) units. The proposed framework is extensively evaluated using the publicly available BRACETS dataset. Experimental results suggest that our ReSTL can accurately derive the BPs from RSs with an average mean absolute error of 0.024±0.011, 0.485±0.118, 0.020±0.011, 0.134±0.068, and 0.031 ± 0.019, respectively for five different tasks. Further these derived BPs can be used for extracting different respiratory vitals, identifying disease conditions efficiently, and retrieving salient breathing cycle information from the RSs.

In this paper, we explore an image denoising method called Total Variation Regularization, wherein we minimise the variation between neighbouring pixels in an image. We use the gradient descent method to minimise the variation, while ensuring that the image does not lose outlines and details. We test various regularization parameters to find an optimum parameter and also test the model on different kinds of images (including those clicked from my own phone).

The flux-barrier rotor is generally agreed to be the optimum topology for SynRM, where design parametrization rules are exhaustive. However, in a specific case where a small diameter of 25.24mm is required, the small number of stator slots and diameter limitation results in a parametrized design with a smaller number of flux barriers. This results in a lower saliency ratio. Design with 2 or more flux barriers must be excluded in this small dimension constraint due to the manufacturing feasibility of small fine blanking. This study evaluates if the conventional parametrization rules for flux-barrier SynRM design is applicable in this condition. 4 alternatives of rotor topologies were proposed and analyzed using FE analysis. The rotors proposed include an innovative segmented rotor as an alternative to the conventional flux-barrier rotors. The results show that the segmented rotor has the highest saliency ratio at 2.07 and a torque density of 2.2e-5 which is comparable to SynRMs in recent literature. It produces a torque of 0.78 Nm, an efficiency of 45.75%, and a power factor of 0.63 at 5000rpm and 50A. The mechanical stress and deformation of the rotor structure are also at an acceptable value of 6.2e5 N/mm 2 and 5.5e-5 mm.

This study conducts a comprehensive analysis of artificial intelligence models' capabilities in solving advanced mathematical problems beyond Calculus II. Notably, the research focuses on evaluating the effectiveness of six optimized models, including a specially designed Physics-Informed Neural Network (PINN), in solving stochastic, linear, nonlinear, and ordinary differential equations. Our findings revealed that the optimized PINN achieved a success rate of 91.03%, outperforming most state-of-the-art artificial intelligence models. This research contributed to the development of artificial intelligence mathematical reasoning, with potential applications to high-level engineering, healthcare, research, sustainability, and big data optimization.

In our research, Calories burnt prediction, we aimed to utilize machine learning to predict calories burnt. The process involves processing the dataset, which contains information about a person. Like every machine learning work, we started with preparing the dataset. We performed exploratory Data Analysis. Moreover, we cleaned the dataset by finding and inputting missing values. Five machine learning models were used: KNN, Decision Tree, AdaBoost, SVM, and XGBoost. The dataset was trained on the above models. We trained and tested with default and hyperparameters to obtain the best result. Some of our models performed remarkably in predicting the calories burnt. We also used Explainable AI to find the link between provided features and our result. The research also highlights the importance of hyperparameters optimization for gaining the best output. All in all, this work demonstrates the introduction of advanced techniques like Machine learning in the physical exercise sector to predict calories loss based on personalized data, which can be further helpful for individual health and fitness programs. After experimenting with all the models using both default and optimized hyperparameter tuning we found that XGBoost provides the best result with RMSE 2.13 and R2 coefficient of 1.

Understanding language is a complex task, both for humans and AI. This paper explores the inherent challenges in language comprehension, focusing on the diculties faced by AI in achieving true context awareness.

AI, especially deep learning algorithm, has proved its potential in wind power prediction; however, the lack of explainability is the main concern to address and this work is the first to investigate the emerging Liquid Neural Network (LNN) to provide necessary transparency in wind power prediction. LNN utilizes the mathematical abstraction of C.elegans and demonstrates liquid/robust behavior in learning and estimation for unseen data. For comparative analysis, the LNN family (i.e., closed form continuous (CfC), Liquid Time Constant) and state-of-the-art recurrent networks (e.g., LSTM and GRU) and 1D-CNN are considered, and the CfC neural network provides the best results on unseen data. CfC models with fully connected layers using only 25 neurons have provided superior results for wind power prediction in different time spans, resolutions, and number of variables.

Reliable oceanic and climate analysis depend on high-quality sensor readings, yet these systems commonly encounter significant sensor limitations, leading to missing data. Addressing this issue is critical for ensuring accurate forecasts and analyses. In this work, the data gap problem is studied by developing physics-regularized machine learning models with multiple-location modeling to forecast missing sensor data. Utilized are recurrent statistical surrogate models that generate hourly 24-hour forecasts. To train these models, we use a selection of five sensor features collected over three years. Introduced is a multi-location modeling scheme that uniquely combines sensor data from nearby buoys as a novel methodology. This approach allows for more stable and accurate predictions compared to forecasting with single buoy data alone. Our experiments reveal that grouping six buoys yields the best forecasting performance. Furthermore, we improve model accuracy by integrating buoy data with numerical ocean models and applying a physicsregularized loss function. This technique mitigates the impact of missing or erratic data, leading to more dependable 24hour forecasts. Our findings demonstrate that the combination of multiple-location modeling and physics-based regularization enhances the stability and accuracy of oceanic data forecasting.

This paper introduces a new theoretical framework for edge detection and pioneers a novel technical route grounded in this framework. In this paper, texture is defined as the disparity in optical properties between two distinct regions. Based on this definition, an XOR operator is devised to precisely extract gradient and orientation information of textures from binary images. By exploiting the bit independence property and the bit difference weighted sum property of binary numbers, the application scope of the XOR operator is expanded, resulting in a new algorithm applicable to all types of image —the Bitwise Texture Extraction Algorithm (BTE). Edge is defined as the texture that delineates the contours of visible objects. Based on this definition, an algorithm named the Bitwise Edge Extraction Algorithm (BEE) is proposed, which is for extracting edge information from the output of BTE. Subsequently, the application of the output of the BEE algorithm, the BEE Map, in instance segmentation tasks is discussed, and the BEE algorithm is optimized by leveraging the linear continuity of edges, enhancing its accuracy in assigning edge affiliations. Finally, this paper tackles the challenge of Edge-to-Noise Separation by defining image noise as all elements in an image that do not constitute target elements. Based on this novel definition, a new approach to noise filtering is introduced: by applying operations that preserve only valid edge clusters for each object within the BEE Map individually, and then combining these outcomes, a noise-free, pure edge map is achieved.

Blockchain technology is a sophisticated mature software that distribute different databases on numerous networked computers at real time. When, there is a variation in databases the software adds an encrypted block of updated data, which keeps the blockchain’s posterity. Banafa_2020 So, blockchain is a technology that help keep data secure and authentic. In that context, social security number is very important and sensitive subject. Social Security Number is the best form of identification there is, because many entities can review and verify person’s identity with this number, and it is always unique to each person. Sorrells_2010 If a person’s Social Security Number has been filched then the persons become helpless because it is very difficult to change Social Security number, as it is linked from educational transcript, professional license to every little aspect in that person’s life. Read_2006

To achieve Artificial intelligence technological singularity(AITS), the most important task is to explain the job for the computer, so that it will understand easily and train the computer ways to accomplish the job successfully and generate and execute the beast result. To accomplish that task, the most important thing to do is to breakdown a big job into bite size pieces, so at the end, combining all that small jobs result will the accomplishment of the big job. In this paper, we will establish a framework that artificial intelligence can break down a big job into small bite size pieces task and at the end, it will combine all the bite size pieces task result and accomplish the big job successfully.

The security of AI systems becomes increasingly paramount, as artificial intelligence (AI) pervades various aspects of our lives. For instance, from autonomous driving to banking and medical diagnostics, AI is playing a pivotal role in powering such security-sensitive systems. However, attacks on these high-risk AI systems could have severe consequences, jeopardizing the lives, finances, and well-being of people. Thus, adherence to fundamental security properties, namely, Confidentiality, Integrity, and Availability (CIA), is essential for AI systems to comply with security standards. In this paper, we first define a simplified abstraction of AI systems to discuss four independent viewpoints: data, models, inputs, and deployment. Subsequently, we conduct a detailed analysis of attack vectors targeting each of these viewpoints, with a rigorous assessment of CIA properties. Proactive identification of attack vectors throughout the entire AI lifecycle is a requisite step toward establishing a secure-by-design framework for AI systems.

Downlink channel temporal prediction is a critical technology in massive multiple-input multiple-output (MIMO) systems. However, existing methods that rely on fixed-step historical sequences significantly limit the accuracy, practicality, and scalability of channel prediction. Recent advances have shown that large language models (LLMs) exhibit strong pattern recognition and reasoning abilities over complex sequences. The challenge lies in effectively aligning wireless communication data with the modalities used in natural language processing to fully harness these capabilities. In this work, we introduce Csi-LLM, a novel LLM-powered downlink channel prediction technique that models variable-step historical sequences. To ensure effective cross-modality application, we align the design and training of Csi-LLM with the processing of natural language tasks, leveraging the LLM's next-token generation capability for predicting the next step in channel state information (CSI). Simulation results demonstrate the effectiveness of this alignment strategy, with Csi-LLM consistently delivering stable performance improvements across various scenarios and showing significant potential in continuous multi-step prediction.

Large Language Models (LLMs), built on the transformer architecture, are revolutionizing optical network management by addressing the complexities of these highly specialized systems. Characterized by real-time performance requirements, multi-vendor equipment interoperability, and intricate signal processing, optical networks face challenges such as managing diverse transmission impairments, including chromatic dispersion, polarization mode dispersion, and nonlinear effects. These challenges demand advanced automation and optimization techniques, which LLMs are well-suited to address. The integration of LLMs provides a scalable and adaptive approach to automating tasks like network configuration, fault diagnosis, alarm management, and routing and spectral assignment (RSA). By enhancing Quality of Transmission (QoT) estimation, optimizing amplifier gain control, and supporting advanced simulation frameworks, LLMs enable efficient and dynamic decision-making. Additionally, LLMs offer user-friendly interfaces and the ability to incorporate Human-in-the-Loop (HITL) systems, ensuring critical decisions are monitored and managed in real-time. The proposed framework for optical networks combines LLMs with digital twin technology, enabling real-time network monitoring, predictive analysis, and scenariobased optimization within virtualized environments. This synergy reduces operational complexity, enhances resource efficiency, and facilitates intelligent, autonomous decision-making. Despite their promise, LLMs face challenges such as hallucinations—producing semantically incorrect or fabricated outputs—and computational latency, particularly in real-time tasks like dynamic reconfiguration and fault resolution. Addressing these challenges requires strategies such as prompt engineering, retrieval-augmented generation (RAG), fine-tuning with domainspecific data, and integrating error prediction models to improve accuracy and reduce risks. Techniques like edge computing, model pruning, and adaptive deployment help mitigate latency concerns while optimizing resource utilization. Energy efficiency is further enhanced through hardware optimization, algorithm refinement, and renewable energy adoption. Studies demonstrate that strategies like lowering graphics grocessing units (GPU) frequencies or employing model parallelism can significantly reduce energy consumption without sacrificing performance. This paper explores the transformative potential of LLMs in optical networks, highlighting their applications in network design, alarm compression, and resource optimization. The results demonstrate that with proper adaptation, including integration with digital twins and sustainable deployment strategies, LLMs can significantly enhance automation, scalability, and energy efficiency. By unifying diverse tasks under a single intelligent framework, LLMs offer a pathway to revolutionize optical network architectures, paving the way for greener, more intelligent, resilient, and adaptive communication systems.

Generative Artificial Intelligence (GenAI) is rapidly being integrated across industries, promising transformative practices in areas such as software development, marketing, healthcare, finance, manufacturing, and customer service, and enhancing efficiency through Human-Artificial Intelligence (HAI) interaction and collaboration. However, one of the major obstacles for fluent HAI interaction is the lack of common context and understanding. Without these, the AI and the user are not guaranteed to be aligned, resulting in inefficient and frustrating collaboration. In this position paper, we propose to integrate model-based understanding of the user into GenAI unlike current use of information retrieval augmented generation in copilots. The model implements hypotheses about human cognition and how human goals, beliefs, and abilities impact behavior and adaptation in interactive environments. This permits the GenAI to infer the latent mental states of the user and predict how the user would adapt to different actions taken by the AI. We emphasize the importance of cognitively informed HAI interaction, advocating for future interdisciplinary research in the intersection of cognitive science and AI, focusing on finding empirical evidence about how GenAIs' ability to infer the latent cognitive states of the user benefits human-GenAI interaction. To facilitate responsible development, implementation and widespread acceptance of cognitively informed AI agents we propose and recommend the CRAFTS AI Framework. Ignoring the principles laid out in the framework could significantly hinder the progress and potential benefits of HAI collaboration.

The article is devoted to frequency sampling filters (FSF's). FSF's are very efficient linear-phase digital filters that can be significantly more computationally efficient than Parks-McClellan FIR filters. With the help of FSF's, it is convenient to implement filter banks that can be used to create efficient high-performance spectrum analyzers. These filters have been known for a long time, but, despite their effectiveness, they are not widely used. The article presents the authors' point of view on the reasons for the weak use of FSF's. It is shown how their transfer function (TF) was obtained using an analog prototype. As an analog prototype, the TF of the filter built with the use of LC-resonators connected in parallel was used. An analysis of the provisions of the theory of synthesis of analog filters was carried out. The TF of the analog filter, which implements the requirements for absolute linearity of the phase response, is considered. The bilinear z-transform method was applied to this TF. An analysis of the frequency characteristics of the digital filter was carried out based on the obtained TF with real coefficients. Applying L'Hôpital's rule, an analytical expression was found for determining the weighting coefficients of the resulting TF. It was concluded that it is necessary to modify this TF in order to achieve frequency independence of the weighting coefficients. The following modification was carried out. The modified TF of the FSF with real coefficients simplifies the processes of approximation and implementation. Studying methods of approximation of an analog prototype facilitates the approach to solving the problem of FSF's approximation. The article provides examples of calculated FSF's and a filter bank based on them. Graphs of frequence magnitude responses and tables of calculated optimal weighting coefficients are shown. The FSF block diagram is shown.