Improving the controllability, portability, and inference speed of diffusion language models (DLMs) is a key challenge in natural language generation. While recent research has shown significant success in complex text generation with language models, the memory and computational power are still very demanding and fall short of expectations, which naturally results in low portability and instability for the models. To mitigate these issues, numerous well-established methods were proposed for neural network quantization. To further enhance their portability of independent deployment as well as improve their stability evaluated by language perplexity, we propose a novel approach called the Quantized Embedding Controllable Diffusion Language Model (QE-CDLM). QE-CDLM builds upon the recent successful controllable DLMs by remodeling the task-specific embedding space via quantization. This leads to a gradient-based controller for the generation tasks, and more stable intermediate latent variables are obtained, which naturally brings in an accelerated convergence as well as better controllability. Additionally, the adaption fine-tuning method is employed to reduce tunable weights. Experimental results on five challenging fine-grained control tasks demonstrate that QE-CDLM compares favorably to existing methods in terms of quality and feasibility, achieving better perplexity and lightweight fine-tuning.

Integrated local multi-energy systems are recognized as a promising option to achieve the ambitious energy and climate goals set by the European Commission for 2030. The nature of integrated systems requires a sound combination of interdisciplinary methodologies and complementary tools. Creating an efficient architecture capable of exploiting synergies between tools is therefore crucial for designing, analysing and operating integrated energy systems. The joint application of different tools needed for analysing a local system can be very demanding and time-consuming due to vastly different data structures and functionalities. To address the need for complementary tools, the aim of this paper is to establish and test an integrated modelling architecture allowing the interaction of tools into a modular toolbox for the optimal planning of integrated local multi-energy systems, and also present key preliminary outcomes.

Lower limb exoskeletons (LLEs) are wearable devices that can restore the movement autonomy of paraplegic users. LLEs can restore the users’ ability to stand upright and walk. However, most of the commercially available and clinically used LLEs rely on the user maintaining balance through the use of crutches. Recent improvements in the design and control of LLEs and other legged robots allow for autonomous balance control. In this work, we implement and evaluate a momentum-based standing balance controller in the Symbitron LLE, consisting of eight active (torque-controlled) and two passive joints. We first investigate how gain tuning of the center of mass tracking control law, part of a multi-task optimal controller, affects balancing performance. We apply pushes on different device locations while in parallel-stance, compare the response for different gains, and derive heuristic guidelines for controller tuning given the control architecture, high-level goals, and hardware limitations. Next, we show how this controller successfully prescribes joint torques to the LLE to maintain balance with a paraplegic user. The LLE can autonomously balance the user and reject mediolateral and anteroposterior pushes in the order of 60 N at hip height (and 40 N at shoulder height) while standing in parallel-stance, staggered-stance with both feet at the same height, and staggered-stance with a height difference of 0.05 m between the feet. This work presents a viable control strategy for torque-controlled light-weight under-actuated LLEs to keep the balance of paraplegic users during stance, which is a necessary starting point towards autonomous balance control during gait.

This paper describes the design and implementation processes of a radio telescope suitable for the detection of the hydrogen line, proposing a receiver system based on Software Defined Radio (SDR). The design and fabrication of the antenna are presented, as well as the results of its characterization (electrical, radiation), both with simulations and measurements in an anechoic chamber. The functionality of the instrument is demonstrated by the satisfactory observations of the hydrogen line emitted by our Galaxy. This is an affordable project that allows the development and application of the skills of telecommunications students since it includes many of the fundamental aspects of electronics applied in this field.

Design patterns are best practices for known problems in a specific context. Many patterns have been proposed in different domains, such as object-orientated programming, software architecture, and workflows, to name a few. However, blockchain interoperability is a recent area of work and, to our knowledge, no design patterns have been defined yet. The purpose of this work was to identify blockchain interoperability patterns that may exist in blockchain interoperability solutions. We identified six patterns through the observation of 35 interoperability solutions. A specification was built for each pattern using the Alexandrian template. The specification was evaluated with five semi-structured interviews with blockchain experts to collect data on the comprehension, completeness, and utility of the patterns. The results show that all interviewees identified the patterns. However, the pattern specification has different degrees of confidence in terms of clarity, completeness, and utility. Finally, all interviewees thought that the proposed patterns may be helpful to software architects in their first blockchain interoperability project.

Kähler forms with associated metrics and cohomology classes are shown taking Kähler–Einstein metric.

This work develops a novel observer-based finite-time methodology to make disturbed wheeled mobile robots (WMRs)  follow a desired reference signal. With the kinematic and reference models, two decoupled systems are computed. Then, each subsystem’s control signal is split into three terms. The first part estimates the disturbances affecting the system, transforming the closed-loop system into two new second-order systems with uniformly bounded and convergent disturbance terms. Then, based on a new sliding surface and a design function, the second and third controller’s parts are employed to accomplish the trajectory tracking task despite disturbances. The proposed controller is compared against a finite-time and a feedback controller. Then, detailed numerical simulations show the outstanding performance of the novel proposal.

Identification of sources of seizures in the brain is of paramount importance, particularly for drug-resistant epilepsy patients who may require surgical operation. Interictal epileptiform discharges (IEDs), which may or may not be frequent, are known to originate from seizure networks. Delayed responses (DRs) to brain electrical stimulation have been recently discovered. If DRs and IEDs come from the same location and the DRs can be accurately localized, there will be a significant step in the identification of the source of seizures. The solution to this important question has been investigated in this paper. For this, we have exploited the morphology of these spike-type, events as well as the variability in their temporal location, to develop new constraints for an adaptive Bayesian beamformer that outperforms the conventional and recently proposed beamformers. This beamformer is applied to an array (a.k.a mat) of cortical EEG electrodes. As the significant outcome of applying this beamformer, it is very likely (if not certain) that the IEDs and DRs for an epileptic subject originate from the same location in the brain. This paves the way for a quick identification of the source(s) of seizure in the brain.

This systematic literature review (SLR) explores the broad application of Artificial Intelligence (AI) in healthcare, with a focus on integrating emotional and cognitive analysis techniques. The review aims to assess the impact of AI technologies on enhancing patient care by analyzing emotional content and supporting patient-centric applications. We examine studies that leverage AI for sentiment analysis, emotion categorization, and predictive modeling using textual data from clinical narratives, patient feedback, and online health discussions. The review identifies key advancements in emotion recognition algorithms, the synergy of cognitive and emotional analysis methods, and the development of AI-powered clinical decision support systems. These applications have significantly improved personalized therapy plans, facilitated early detection of mental health disorders, and enhanced support for clinical interventions. Despite these developments, challenges such as ethical AI implementation, patient confidentiality, and algorithmic biases remain. Nevertheless, AI's potential to transform healthcare practices is evident, promising a future of more informed, efficient, empathetic, and patient-centered healthcare. This study highlights AI's transformative impact on healthcare by providing a comprehensive overview of its role in emotional and cognitive analysis, identifying emerging trends, and advocating for the integration of AI capabilities with the human aspects of healthcare.

Transformer requires a fixed number of layers and heads which makes them inflexible to the complexity of individual samples and expensive in training and inference. To address this, we propose a sample-based Dynamic Hierarchical Transformer (DHT) model whose layers and heads can be dynamically configured with single data samples via solving contextual bandit problems. To determine the number of layers and heads, we use the Uniform Confidence Bound algorithm while we deploy combinatorial Thompson Sampling in order to select specific head combinations given their number. Different from previous work that focuses on compressing trained networks for inference only, DHT is not only advantageous for adaptively optimizing the underlying network architecture during training but also has a flexible network for efficient inference. To the best of our knowledge, this is the first comprehensive data-driven dynamic transformer without any additional auxiliary neural networks that implement the dynamic system. According to the experiment results, we achieve up to 74\% computational savings for both training and inference with a minimal loss of accuracy.

Ultrasonic attenuation can be used to characterize tissue properties of the human breast. Both quantitative ultrasound (QUS) and ultrasound tomography (USCT) can provide attenuation estimation. However, limitations have been identified for both approaches. In QUS, the generation of attenuation maps involves separating the whole image into different data blocks. The optimal size of the data block is around 15 to 30 pulse lengths, which dramatically decreases the spatial resolution for attenuation imaging. In USCT, the attenuation is often estimated with a full wave inversion (FWI) method, which is affected by background noise. In order to achieve a high resolution attenuation image with low variance, a deep learning (DL) based method is proposed. In the approach, RF data from 60 angle views from the QTI Breast Acoustic CT TM Scanner are acquired as the input and attenuation images as the output. To improve image quality for the DL method, the spatial correlation between speed of sound (SOS) and attenuation were used as a constraint in the model. The results indicate that by including the SOS structural information, the performance of the model is better. With a higher spatial resolution attenuation image, further segmentation of the breast can be achieved. Segmentation of breast using the DL-based attenuation image had a high spatial correlation with the segmentation from the FWI-based SOS image, i.e., with correlation coefficients greater than 0.64, and Intersection over Union of the ROI greater than 0.73, which confirms that the DL-generated attenuation images can be used to do image segmentation.

System level simulation of neuro-memristive circuits under variability are complex and follow a black-box neural network approach. In realistic hardware, they are often difficult to cross-check for accuracy and reproducible results. The accurate memristor model prediction becomes critical to decipher the overall the circuit function in wide range of non-ideal and practical conditions. In most neuro-memristive systems, crossbar configuration is essential for implementing multiply and accumulate calculations, that forms the primary unit for neural network implementations. Predicting the specific memristor model that best fits the crossbar simulations to make it explainable is an open challenge that is solved in this paper. As the size of the crossbar increases the cross-validation becomes an even more challenging. This paper proposes predicting the memristor device under test by automatically evaluating the I-V behavior using Random forest and Extreme Gradient Boosting algorithms. Starting with a single memristor model, the prediction approach is extended to memristor crossbarbased circuits explainable. The performance of both algorithms is analysed based on precision, recall, f1-score and support. The accuracy, macro average and weighted average of both algorithms at different operational frequencies are explored.

As machine learning is being used in numerous applications, there is even more concern regarding algorithms within artificial intelligence. This paper focuses on analyzing the biases incorporated in a setup like FOLD-R++. The findings hold significance to both academia and industry in establishing what a fair and neutral machine learning algorithm is. The article contains an exhaustive literature review about biases in machine learning and a detailed explication. The chapter re-examines previous studies on the efficacy of this algorithm, unearthing the limitations in earlier literature and urging more research.This involves purposefully choosing a dataset from Kaggle, metrics applied in evaluating the algorithm, and a detailed experimental design. The results of these tests of different test scenarios have displayed that the algorithm is correct but vulnerable to bias and efficient across different domains. The results are discussed, and comparisons with what is available in the table regarding system-biased algorithms are made. In conclusion, this study contributes to the existing literature on machine learning and highlights certain shortcomings concerning the use of the FOLD-R++ algorithm. Thus, it demonstrates the need to address issues of unexplained inequality and develop reliable algorithms for decision-making. The study acknowledges these shortcomings to the extent that it becomes a vital stage toward developing more justified machine learning.

The integration of artificial intelligence (AI) into energy networks significantly advanced short-term forecasting, particularly in smart meter applications. However, as distributed energy resources proliferated and energy systems grew in complexity, traditional centralized approaches to data analysis became insufficient in addressing privacy-preserving challenges. Federated learning (FL) emerged as a promising solution, leveraging distributed data sources while safeguarding user privacy. Nonetheless, FL encountered inherent vulnerabilities to adversarial attacks during model training, undermining its reliability and effectiveness. Existing techniques to eliminate these attacks often required additional frameworks for detection, imposing an added burden on devices. To address this issue, we proposed a novel method called federated random layer aggregation (FedRLA). It aggregated only one randomly chosen neural network layer on the server in a privacy-aware manner, leaving the remaining layers unchanged. FedRLA exhibited superior resilience against adversarial attacks by confining attackers to a single neural network layer. Our simulations, focusing on household energy consumption, demonstrated that FedRLA achieved 3.56 times less data transmission compared to FedAvg during global model training. This enhanced efficiency translated to improved energy usage and resource conservation. Furthermore, FedRLA performed better in the presence of differential privacy under attack and no attack conditions.

Area and power-efficient hardware implementations are crucial for the widespread adoption of post-quantum cryptography (PQC) algorithms like FALCON. One of the main operations in FALCON is the Number Theoretic Transform (NTT), which needs to be performed with many prime numbers. Having one set of twiddle factors (TFs) for each prime makes storing all of these TFs impractical. In this paper, we propose an architecture for generating TFs on the fly for FALCON-oriented NTT, designed for area and power efficiency. Our approach dynamically generates TFs during NTT computations, significantly reducing on-chip memory requirements. The ASIC implementation results demonstrate significant improvements, with the proposed design reducing on-chip memory requirements by 99%, occupying 95% less area, and consuming 87.4% less power compared to the traditional ROM-based implementation. Furthermore, our design achieved a much higher maximum clock frequency, indicating superior performance in accessing twiddle factors. These findings highlight the potential of our proposed architecture for efficient hardware implementations of FALCON-based cryptographic systems.

Rapid changes arising from the technological revolution, mainly the advent of artificial intelligence, have been defining moments of the 21st century. Education and employment stand at the forefront of AI disruption; as a result, academia and industry, as the principal agents of this development, are deemed mutually involved in mitigating risks. Nonetheless, triggered by massive technological, economic, and social development at an accelerating rate, the industry-academia gap in orientation and direction is getting wider, leading to detrimental disarray. As the disruptive impact has far-reaching consequences, the role of government entities as the regulatory body and civil society as end users of these technologies and their impact on the environment is relevant and imperative. The extensive work on the quintuple helix model of innovation has been theorizing innovation ecosystems and processes without addressing the pragmatic aspect of stimulating innovation for specific problems. This paper put forward a vision for mitigating AI disruption by leveraging value creation and exchange harnessing symbiotic aspects of the helices by considering the necessity of knowledge circulation to address environmental and social impact. Finally, it advocates a series of mitigating strategies through a collaborative council empowered by human-centric ICT-Platform that can transcend to a dynamic policy-making mechanism.

Onboard intelligent processing is widely applied in emergency tasks in the field of remote sensing. However, it is predominantly confined to an individual platform with a limited observation range as well as susceptibility to interference, resulting in limited accuracy. Considering the current state of multi-platform collaborative observation, this article innovatively presents a distributed collaborative perception network called DCP-Net. Firstly, the proposed DCP-Net helps members to enhance perception performance by integrating features from other platforms. Secondly, a self-mutual information match module is proposed to identify collaboration opportunities and select suitable partners, prioritizing critical collaborative features and reducing redundant transmission cost. Thirdly, a related feature fusion module is designed to address the misalignment between local and collaborative features, improving the quality of fused features for the downstream task. We conduct extensive experiments and visualization analyses using three semantic segmentation datasets, including Potsdam, iSAID and DFC23. The results demonstrate that DCP-Net outperforms the existing methods comprehensively, improving mIoU by 2.61%~16.89% at the highest collaboration efficiency, which promotes the performance to a state-of-the-art level.

With the advent of Healthcare 4.0, there is increased interest from researchers the world over in the application of modern, cutting-edge Artificial Intelligence (AI) and Quantum Artificial Intelligence (QAI) algorithms in solving healthcare challenges. The era of Quantum Computing (QC) promises to bring significant advancements in several areas of healthcare such that it may be sensible to give this hybrid Quantum/Classical paradigm its own name-Healthcare4Q. The potential of QC will extend the reach of Healthcare4Q with the help of diverse technologies such as quantum-enabled wearables, quantum-secure transfer and storage of data, and quantum computing at edge, fog, and cloud. All of these technologies promise to catapult Healthcare4Q to become the most capable healthcare framework in the advancement of medical innovations and improvement of patient care. An integral part of a person's health lies in cardiovascular health, and thus prioritizing and optimizing cardiovascular health remains vital to the broader goals of public health and healthcare sustainability. In this study, under the paradigm of Healthcare4Q, we propose a framework called the Quantum AIdriven Heart Health Framework (QAIHHF) that can provide advanced predictive intelligence to healthcare providers by utilizing historical and real-time data and processing capabilities proposed in Healthcare4Q. We show that when applied to various diagnostics and health indicators such as ECG data, the Quantum AI provides accuracy at a level equal to or higher as compared to the classical methods thus proving itself to be the critical component that will herald the era of Healthcare4Q.