Do Large Language Models have cognitive abilities? Do Large Language Models have understanding? Is the correct recognition of verbal contexts or visual objects, based on pre-learning on a large training dataset, a manifestation of the ability to solve cognitive tasks? Or is any LLM just a statistical approximator that compiles averaged texts from its huge dataset close to the specified prompts? The answers to these questions require rigorous formal definitions of the cognitive concepts of "knowledge", "understanding" and related terms.

This study endeavors to establish an simple, robust, and transparent forecasting framework that is easily deployable, computationally efficient, and adaptable across the various sectors constituting distributed integrated energy systems. Accordingly, a probabilistic day-ahead forecasting framework and its application are presented. A personalized standard load profile (PSLP) is then applied to electricity and heat demand, as well as PV generation. Furthermore, it is extended to incorporate a quantile personalized standard load profile employing quantile regression. Using a PCHIP interpolation, the empirical cumulative distribution function is approximated from the non-parametric approach and the probability density function derived. Furthermore, it is shown how discrete convolution can be used to determine the joint probability density function of the distributed net load. A case study was conducted at a commercial logistics facility that utilized a decentralized energy system to showcase the forecasting framework and its application for optimization in the presence of uncertainty. This involved using the net load and its probability density functions for uncertainty-based optimization.

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.

Exploring the fundamental principles of system modeling in electrical engineering, this study delves into the transformative power of the d-q transformation, highlighting its pivotal role in rendering time-varying systems into a coherent steady-state representation. Departing from conventional approaches, the study carefully navigates the complexities of single-phase transformer configurations, utilizing the Clarke and Park transformations to seamlessly transition between electrical coordinates and the d-q frame. Through extensive derivations, dynamic equations are formulated in both α-β and d-q coordinates, providing a detailed understanding of system dynamics under specific loads. In addition, the study extends the analysis to a generalized multi-winding transformer model that accommodates a wide range of transformer setups. With detailed mathematical derivations, insightful visual aids, and clear state-space representations, this work attempt to be a resource for researchers, engineers, and practitioners seeking to unravel the intricacies of electrical system modeling and analysis.

We present a Multiple Input Multiple Output fuzzy path tracking controller that reduced its rule-base using Hierarchical Rule Base Reduction. A stability proof and outdoor results are also presented.

Phase aberration in ultrasound images is caused by inaccurate information in the sound speed distribution in the medium and can result in image distortion, such as shape change and position shifting of the imaged objects. Various methods, including cross-correlation-based methods, have been applied to the distorted images to estimate phase aberration. In this paper, we first propose that the position shifting induced by the phase aberration causes the estimated phase aberration to be inaccurate. Then, we propose an equation relating the estimated phase aberration to the true one and the equations to predict position shifting. Finally, we present a forward model for estimating phase aberration. Considering phase aberration as a function of the array element position, the theory shows that both the constant term and linear term of the true phase aberration will be cancelled in the estimated phase aberration as they result solely in predictable position shifting. Field II simulations and data from tissue-mimicking phantom were used to validate the proposed theory. The theory was also applied to improve the estimation of the initial delay of ultrasound probes in both Field II simulations and experimental phantom study. Other potential applications of the proposed theory were also discussed.

Power electronics embed intelligence into energy systems. Modern artificial intelligence systems place massive demand on electrical power, and in return, introduce new opportunities in improving energy systems design and implementation. Power electronics systems demand extreme performance and sophisticated functions. Innovations are needed from new and better semiconductors, novel passive components and devices, and more complex system modeling and control. Modern machine learning tools can learn highly complex physical characteristics and control strategies of future power electronics systems, given sufficient training data and computational power. Generative artificial intelligence are rapidly developing. Structured, cognitive understanding of abstracted human intelligence is in demand. Power "The strive for more power is infinite, and the strive for more intelligence is also infinite, leading to infinite entanglement between artificial intelligence and power electronics systems." "We can do it, if we choose to do it."

This paper explores the problems caused by spam messages on the internet. It talks about different types of spam, like email spam and fake advertisements on social media. The paper also looks at how spam can make things difficult for people, like draining resources and tricking them into clicking on harmful links. The paper suggests fighting against spam by enabling filters in emails, using third-party spam filtering, avoiding clicking suspicious links, reporting, and blocking and avoiding using personal or professional emails. Lastly, the paper discusses what might happen with spam in the future and the challenges we may face. Overall, the paper aims to help and understand then find ways to deal with it for a safer and smoother experience.

We experimentally show novel phenomena: (1) in the same double slit or cross double slit experiment, within macroscopic distances from diaphragm, the patterns are non-interference and gradually evolve to the interference patterns near the screen;

Robotic assembly tasks are open challenges due to the long task horizon and complex part relations. Behavior trees (BTs) are increasingly used in robot task planning for their modularity and flexibility, but creating them can be effortintensive. Large language models (LLMs) have recently been applied in robotic task planning for generating action sequences, but their ability to generate BTs has not been investigated. To this end, We propose LLM as BT-planner, a novel framework to leverage LLMs for BT generation in robotic assembly task planning. Four in-context learning methods are introduced to utilize the generative and natural language processing capabilities of LLMs to produce task plans in BT format, reducing manual effort and ensuring robustness and comprehensibility. We also evaluate the performance of fine-tuned fewer-parameter LLMs on the same tasks. Experiments in simulated and real-world settings show that our framework enhances LLMs' performance in BT generation, improving success rates in BT generation through in-context learning and supervised fine-tuning.

A document by Hani Rabiee. Click on the document to view its contents.

IdentifiEye is designed to enhance the social interaction capabilities of individuals affected by prosopagnosia, commonly known as face blindness. Affecting exactly 3.08% of the American population (300 million people globally), prosopagnosia impedes the ability to recognize faces, leading to social challenges. IdentifiEye glasses utilize cutting-edge facial recognition and augmented reality technologies to help users identify faces, thereby improving their social confidence and ability to form meaningful connections. Within this project, I implemented the use of various face recognition API, detection models, as well as back-end storage system which allowed for the development of the application and glasses that are being used hand-in-hand. More specifically, this application will scan and save a photo and important tags, information, or voice memos into the AWS backend system which can be accessed in any future situations or interactions to remember for future use. The face recognition software that was trained and tested within the application has a cloud storage and recognition system which achieves over 99% in face recognition accuracy. Creating this application that is tied within glasses will allow for a significant increase for memorable social interactions that will change the lives of hundreds of millions of people globally.

The lack of explainability using relevant clinical knowledge hinders the adoption of Artificial Intelligence-powered analysis of unstructured clinical dialogue. A wealth of relevant, untapped Mental Health (MH) data is available in online communities, providing the opportunity to address the explainability problem with substantial potential impact as a screening tool for both online and offline applications. We develop a method to enhance attention in popular transformer models and generate clinician-understandable explanations for classification by incorporating external clinical knowledge. Inspired by how clinicians rely on their expertise when interacting with patients, we leverage relevant clinical knowledge to model patient inputs, providing meaningful explanations for classification. This will save manual review time and engender trust. We develop such a system in the context of MH using clinical practice guidelines (CPG) for diagnosing depression, a mental health disorder of global concern. We propose an application-specific language model called ProcesS knowledge-infused cross ATtention (PSAT), which incorporates CPGs when computing attention. Through rigorous evaluation on three expert-curated datasets related to depression, we demonstrate application-relevant explainability of PSAT. PSAT also surpasses the performance of nine baseline models and can provide explanations where other baselines fall short. We transform a CPG resource focused on depression, such as the Patient Health Questionnaire (e.g. PHQ-9) and related questions, into a machine-readable ontology using SNOMED-CT. With this resource, PSAT enhances the ability of models like GPT-3.5 to generate application-relevant explanations.

5G cellular systems are slowly being deployed worldwide delivering the promised unprecedented levels of throughput and latency to hundreds of millions of users. At such scales security is crucial, and consequently, the 5G standard includes a new series of features to improve the security of its predecessors (i.e., 3G and 4G). In this work, we evaluate the security of currently deployed 5G commercial networks in Europe and North America. Specifically, by collecting 5G signaling traffic in the wild in several cities in Spain, Germany, France, Canada, and the USA, we i) fact-check which 5G security enhancements are implemented in current deployments, ii) provide a rich overview of the implementation status of each 5G security feature in a wide range of 5G commercial networks in Europe and North America and compare it with previous results in China, iii) analyze the implications of optional features not being deployed, and iv) discuss on the still remaining 4G-inherited vulnerabilities. Our findings indicate that the rollout of 5G security features in commercial networks is still a work in progress. On the one hand, many networks continue to rely on 4G for their core network operations, which hinders the deployment of new security features (e.g., SUCI) and, on the other hand, fully-fledged 5G deployments lack mandatory security measures such as GUTI reallocation after paging. Moreover, we find that some operators fail to provide proper temporary identifier randomization, in both 4G and 5G networks. Some of the obtained results are aligned with results previously reported from China [1] and keep the European and North American studied networks vulnerable to some 4G attacks, during their migration period from 4G to 5G. Conversely, networks deployed in North America exhibit stronger adherence to 5G security standards, with near-complete compliance observed, in contrast to deployments in China and Europe where compliance levels are comparatively lower.

The 3D point cloud plays a crucial role in deep learning-based vision tasks by providing precise spatial and depth information, leading to its increasing importance in various applications. However, the sparse nature of 3D point clouds poses computational challenges. Researchers have explored the Submanifold Sparse Convolutional Network (SSCN) for processing point cloud data while preserving sparsity. Nevertheless, existing Convolutional Neural Network (CNN) accelerators encounter difficulties in effectively handling SSCNs, prompting recent studies to focus on developing dedicated accelerators for point cloud networks to improve processing performance. This brief presents a specialized hardware architecture designed for SSCNs to address the challenges of effectively processing sparse 3D point clouds. The proposed accelerator achieves a significant 2.51× improvement in throughput density compared to previous works, highlighting its effectiveness in point cloud processing.

Cybersecurity knowledge graphs (CKGs) integrate diverse sources of cyber threat intelligence (CTI) into a structured, queryable format, offering scalable solutions for automating proactive and real-time security responses. This has led to their increased adoption, improving the workflow and effectiveness of security professionals. However, constructing CKGs requires extracting entity-relation triples from unstructured reports, a process complicated by report complexity, specialized cybersecurity terminologies, and language ambiguity. As a result, existing pipeline approaches suffer from error propagation, resulting in low accuracy and poor generalizability. This paper introduces the Context-aware Threat Intelligence Knowledge Graph (CTiKG) framework, designed to accurately identify and classify threat entities and their relationships from CTI reports. CTiKG integrates hybrid NLP models that leverage SecureBERT embeddings and expert knowledge encoded in an ontology to control classification errors and reduce error propagation, thereby enhancing extraction accuracy. Evaluations using the augmented DNRTI-STIX2 dataset, which features 21 entity categories adhering to STIX-2.1 standards, demonstrate the model's superior performance compared to state-of-the-art methods, with increases of 2-3% in NER and up to 5% in RE in terms of precision, recall, and F1-scores. Further validation on the DNRTI and STUCCO datasets and a practical cybersecurity use case illustrate the framework's robustness. The DNRTI-STIX2 and curated CTI datasets are available on GitHub to support further research.

This paper introduces a new softwareoriented collaborative project-based learning approach to biomedical signal processing (BSP) for graduate biomedical engineering (BE) students. The course simulates industry-like practices under an imposed work environment and acquaints the participants with biomedical signal evaluation and quality assurance procedures required for software standardization. Background: Most academic approaches to teaching BSP focus either on acquisition procedures of electrophysiological data, predefined menu-driven signal processing methods using virtual laboratories, or ordinary SP procedures applied to biomedical signals without reaching the nature of the data and reflecting the future workplace. Methodology: The BE students individually code BSP procedures from scratch under teacher supervision, verify the results against the references and prepare reflective reports from their investigations. Students are involved in a simulated industry-like collaborative electrocardiogram signal processing project led by a project manager (PM) and assisted by a software architect (SA) and project coordinator (PC), with imposed working environment and standardization procedures. Research Questions: 1) Does implementing the BSP algorithms from scratch enable one to understand their nature thoroughly? 2) Does the simulated industry-like BSP course improve software engineering and soft skills via the hidden poll methodology? 3) Does including the PM, SA, and PC in the pipeline-dependent collaborative project positively result in its completion? The paper hypothesises that learning the BSP through a proposed project-based simulated industry-like approach with imposed work conditions improves understanding of BSP principles, computer programming skills and social competencies in developing a collaborative project. Findings: The participants significantly enriched BSP-related knowledge after the course, improved computer programming skills (p < .0001; non-parametric Wilcoxon Signed-Rank test), and enhanced soft skills in collaborative work (p < .001) and public presentations (p < .001). The course participants valued the role of algorithm prototyping stages and QA procedures according to official standardization rules. The PM and SA enabled the smooth software development process, while the PC has proven helpful in resolving intrinsic conflicts.

There is a significant interest in Artificial Intelligence (AI) systems for automating decision making. While there are several AI technology solutions to choose from, there is a need to understand the fundamentals of human decision making and conceptualizing required AI system architecture. However, the challenge is: how can we conceptualize the design of an AI system architecture? This paper proposes trimodal thinking for architecting human-centric AI systems. This paper reviews dual system thinking approach, fast and slow, from the well-known book, published by Daniel Kahneman, with a specific purpose to conceptualize the human-centric AI system architecture. This review found that the fast intuitive system 1 and slow self-controlled deliberate system 2 are useful for conceptualizing the design of the AI system architecture. However, there is a need for system 3 to adjust, balance, connect, regulate and switch the control between these two systems. System 3 is critical for mitigating system 1 unconscious and system 2 conscious biases, power imbalance and noise. Relevant concepts, from principle-agent theory, homeostasis, control theory, and the atomic human, were also used to guide the proposal of trimodal thinking. Thus, this paper proposed trimodal thinking and demonstrated its use for architecting human-centric AI systems using the combination of (1) fast, (2) slow and (3) control systems. This paper also highlighted several implications for broader enterprise architecture discipline and indicated new topics across decision science and computer science areas for further research. It is anticipated that this work will provide a necessary foundation for conceptualizing technology agnostic and human-centric AI systems architecture. This will support the complex AI system explainability and transparency needs of non-technical users and technical AI system engineers.