In this paper, an analysis and performance review of a unique hybrid high-power lithium-iron phosphate cell (HP-LFP) with a high cycle life and fast charge/discharge rate is presented. The new hybrid cell has been developed under the framework of the EU-funded project Hybrid Energy Storage Station (HEROES). The proposed HP-LFP cell simultaneously achieves an optimized energy and power throughput making it useful for applications where both power and energy are demanded such as different types of electric vehicles (EVs), microgrids, and charge stations. The cathode of the hybrid HP-LFP cell is similar to the LFP batteries while the supercapacitorlike anode is based on hard carbon. Thus, the HP-LFP cell has some similarities to the recent lithium-ion capacitors (LiCs) but it behaves differently than the existing LFP batteries and supercapacitors. To distinguish the differences, HP-LFP cell samples are tested, characterized, and analyzed to review their performance compared to the existing technologies. Different standard tests including the open-circuit test, hybrid pulse power characteristics test, and capacity test are fulfilled. The test data is used for model parameterization and parameter sensitivity analysis with respect to the cell's state of charge, temperature, and charge rate. The cell performance has been validated through mission profile testing in an EV fast charge station. Based on the analysis, the paper provides a comparison between different energy storage cells.

Grating-based dark-field and phase contrast x-ray imaging seeks to enhance conventional attenuation based x-ray imaging by improving soft-tissue contrast and revealing additional microstructural details. Initial studies have shown that the dark-field channel in particular significantly complements and enhances conventional thoracic radiography. The technique uses a three-grating interferometer in the x-ray beam, which modulates the wavefront and creates an interference pattern on the detector. One major challenge in adapting grating-based dark-field computed tomography (CT) for medical applications is to keep the data acquisition time within some seconds. An ideal setup involves a continuously moving source/detector arrangement, but this approach precludes conventional phase retrieval and stepping techniques that require multiple projections acquired at the same rotation angle with different grating positions. In this work, we introduce a new algorithm that enhances phase retrieval for continuously acquired data and reduces movement-based cross-talk artifacts. Unlike the sliding window phase retrieval method, which performs phase retrieval locally in patches, our algorithm conducts a global analysis of the entire sinogram using Tikhonov regularization. We validate the effectiveness of this algorithm with experimental data from a human-scale dark-field CT prototype, demonstrating its superiority over the patch-wise sliding window method and its ability to produce artifact-free reconstructions.

State estimation based on Kalman Filter (KF) has been widely considered for state-of-charge (SoC) prediction of lithium-ion (Li-ion) batteries. However, the KF is a model-based approach, and its performance declines when faced with modeling inaccuracies resulting from the nonlinear and timevarying nature of Li-ion batteries. To tackle the modeling uncertainties, a novel hybrid variant of KF is deployed, in which priori estimations are attained similar to the KF algorithm while the posterior estimations are obtained using a recurrent neural network (RNN) integrated into the KF architecture. The proposed method yields enhanced robustness in maintaining the accuracy of SoC estimations in the presence of model mismatches, e.g. due to the aging of batteries. In the proposed method, named KalmanNet, the RNN learns from battery data to predict the optimum Kalman gains for the minimization of SoC estimation error. The proposed method is trained and tested using actual battery data of a high-capacity pouch Li-ion cell replicating real-life driving cycles of electric vehicles. The significance of the results is twofold: 1-When exposed to a model mismatch of ±5%, the SoC estimation accuracy is improved by about 0.5% compared to EKF. 2-The size of required training data is reduced by 20% compared to similar machine learning algorithms.

In recent years, deep learning has significantly advanced the field of medical image recognition, enabling accurate diagnosis and prognosis prediction from complex image data. This paper proposes a novel convolutional neural network (CNN) architecture tailored for medical image recognition tasks, such as disease detection and tissue classification. Our framework incorporates residual connections and attention mechanisms, enhancing feature extraction and representation. The model is evaluated on two standard datasets: chest X-rays for pneumonia detection and histopathological images for tumor classification. The results demonstrate state-of-the-art performance, with accuracy and sensitivity outperforming baseline methods. The proposed method offers a robust solution for automated medical image analysis, with potential applications across various domains in healthcare.

MmWave MIMO (millimeter wave multiple-input multiple-output) systems are critical for enabling the efficient use of spectrum and ultra-fast transmission speeds required by 5G and beyond 5G wireless networks. However, accurate channelestimation over mmWave MIMO systems is challenging due to presence of impulsive noise due to environmental factors, which significantly degrades performance of classical Bussgang/meansquared based channel-estimation methods. To mitigate signal impairments due to unknown non-Gaussian noise processes, this work proposes novel information theoretic learning (ITL) based sparse channel estimation algorithms, namely, zero attracting MCC, and Hyperparameter-Free zero-attracting MCC (ZAMCC). These ITL based algorithms exploit the sparse nature of mmWave MIMO channels and mitigate the adverse effects of impulsive noise. Computer simulations are presented for the performance evaluation for the proposed ITL based algorithms assuming realistic mmWave MIMO scenarios. From the simulations, it is inferred that the proposed ITL based algorithms are promising for accurate hyperparameter-free channel-estimation for practical mmWave MIMO channels impaired by unknown non-Gaussian noises.

While infrastructural development is still growing tremendously, immediate concern is on a responsive road network prediction framework for quickevolving environmental and social demands. In the conventional road planning methodology, the heavy reliance on time consuming and laborintensive manual surveys and data analysis in urbanization processes that are fastgrowing has made it increasingly inefficient. After extensive exploration, this paper presents a very innovative application based on the architecture of UNet, which utilizes geospatial data to predict road networks that could actually exist. In training the model, river networks, residential area data, and elevation data were used. Using Focal Loss with a pos_weight tensor, the model will care more about smaller roads and remote residential areas than before; these may be underrepresented when considering complex urban and rural landscapes due to class imbalance. Results have shown that while the model is limited with the data sources and computational power of the day and hence predicts large roads relatively well, it still demonstrates the full potential an AI model has in road network development and expansion. It improves regional connectivity that can help boost economic growth as well as contribute toward strategic smart city planning. This study represents significant development in road network prediction technology and especially points out its importance in improving the economic and social infrastructure by optimized road network construction.

Gut microbiome research has made tremendous progress, especially with the integration of machine learning and artificial intelligence that can provide new insights from complex microbiome data and its impact on human health. The use of explainable artificial intelligence is becoming critical in medicine, and adopting it in gut microbiome prediction models is appealing for providing more transparent and trustworthy models in clinical research. This scoping review evaluates the use of machine learning and explainable artificial intelligence techniques and identifies existing gaps in knowledge in this research area to suggest future research directions. A systematic search was conducted in PubMed and Scopus to include 59 articles published between 2018 and 2023. Different clinical applications of machine learning and artificial intelligence techniques in gut microbiome studies were applied in the reviewed articles, with Random Forest as the most popular and the most used algorithm, and a large prevalence of black box non explainable models. Finally, not enough attention was paid to the reproducibility of the research work published. This review presents the opportunities that remain to advance research regarding the explainability of artificial intelligence models in the field of microbiome, supporting and accelerating the application of microbiomebased precision medicine in the future.

The Si insulated-gate bipolar transistor (IGBT) and SiC MOSFET hybrid switch offers an effective approach to achieve both high efficiency and cost-effectiveness. Given the varying operational conditions of VVVF (Variable Voltage Variable Frequency) inverters, adaptive control of delay-time and frequency division is crucial to optimizing performance. However, traditional optimization methods are computationally expensive which requires too long time to complete optimization, making them impractical for use in VVVF inverters in vehicles such as trains or cars. This paper introduces a novel approach that divides the optimization process into offline and online stages, employing a Deep Learning approach utilizing Neural Net (NN) algorithm in both stages. This method significantly reduces the time required for optimization compared to conventional adaptive techniques, enabling real-time applicability in VVVF inverters for vehicles. The proposed method is validated using simulation data to train the NN model and test its accuracy, which can be further evaluated with real-world data in future studies.

The Internet of Bio-Nano Things is a promising technology to reduce turnover and increase resolution of medical data collection. Therefore, tiny devices, so-called Nano Machines (NMs) are placed inside the human body, where the collect, process and transmit information to devices outside the human body. Interfaces between in-body and the external electrical domain are crucial to the realization of this technology. In this paper, we propose sweat glands, a pre-existing infrastructure, as a novel interface through which NMs can transmit information from the inside to the outside of the human body. We provide a detailed mathematical model of the envisioned communication system and evaluate its communication performance using different types of detectors (e.g., neural network based detector).

Increasing model complexity has introduced significant challenges in ensuring the safety and ethical alignment of autonomous systems responsible for generating language outputs. Human-centric approaches to safety evaluation, while effective in isolated cases, struggle to scale alongside the exponential growth of model parameters and data requirements. A novel automated pipeline for dynamic data creation is introduced, capable of generating adversarial inputs that expose vulnerabilities in model behavior and allow for iterative refinement. This method eliminates the need for human oversight, offering a more scalable and efficient mechanism for ensuring alignment with predefined safety metrics. Experiments with the LLaMA model demonstrate that dynamic data creation can significantly reduce unsafe outputs, improve response consistency, and enhance the model's robustness through continuous feedback. The findings highlight the transformative potential of automated data generation in addressing the limitations of human-dependent safety evaluation, making it an invaluable tool in the broader field of model alignment and testing. Through this automated approach, safety testing becomes more efficient, cost-effective, and consistent, enabling broader applications across diverse contexts.

A document by Alejandro de la Fuente . Click on the document to view its contents.

This paper addresses the problem of achieving any desired group consensus in multi-agent systems (MAS) operating in arbitrary signed networks. By focusing on arbitrary network topologies, we develop a novel control strategy that drives the agents' states toward any desired state, independent of initial conditions or network topology. A key contribution is the introduction of a switching control mechanism that expands the range of reachable states, ensuring that any targeted outcome is attained for a given initial condition without any constraints on the network structure. Our approach not only guarantees the formation of desired clusters but also ensures that each agent converges to the specified consensus value within its cluster. Mathematical analysis establishes conditions for achieving system stability and reachability to any desired state, based on which the control input is designed. Finally, the theoretical results are validated through extensive simulation results.

The coordination of muscles in human locomotion is commonly understood as the integration of motor modules known as muscle synergies. Recent research has delved into the adaptation of muscle synergies during the acquisition of new motor skills. However, the precise interplay between modulated muscle synergies during movement according to motion requirements remains unclear. Here, we aim to elucidate the alterations in locomotor synergies across various lower-limb motion strategies and motor tasks. Our findings reveal consistent weights of muscles in muscle synergies alongside varying timing activation aligned with specific motion requirements. It shows that spatial muscle synergies remain stable across different motor tasks, but humans adjusted the timing activation of these modules (temporal muscle synergies) to meet the motor requirements. To classify temporal muscle synergies and quantify connection weights for both self-connections and connections between muscle synergies, we employed a graph neural network. Our results demonstrate that muscle synergy 4, responsible for elevating the thigh to propel forward during the swing phase, experiences pronounced enhancement with changes in motion strategies. Furthermore, we observed a reduction in the self-connection of muscle synergy 2, implicated in stabilizing body posture, during motion tasks other than normal walking. Additionally, the connections between muscle synergy 2 and other synergies diminished, indicating more adaptation in muscle synergy 2 to achieve stabilization in more challenging motor tasks. The validity of these findings was verified through five-fold cross-validation, affirming the efficacy of our approach in elucidating neuro-modulation mechanisms in human locomotion. Our proposed methodology holds promising implications for the development of personalized training strategies, offering insights into the intricate interactions among different muscle synergies in accomplishing motor tasks.

The significance of collaboration and coordination in these systems cannot be overstated. They are fundamental to ensuring that individual agents, each with their unique capabilities and roles, can function together effectively. This paper delves into the theoretical foundations and practical mechanisms of collaboration and coordination within MAS, exploring their evolution, key concepts, and real-world applications.

The binding of proteins and ligands is critically important in drug design. However, accurately and efficiently predicting binding affinity still remains a formidable challenge. Recent research has established that deep learning-based computational methods provide a more straightforward and effective solution for quantifying binding affinity compared to traditional experimental approaches. In this study, we introduce a novel sequence-based deep learning architecture, termed ALSRI-Net (Attention-based Long-Short-Range Interaction Network). ALSRI-Net employs self-attention and cross-attention mechanisms to distinctly capture long-range and short-range interaction features between proteins and ligands. By integrating these interaction features, ALSRI-Net delivers more precise and robust predictions. The performance of ALSRI-Net was rigorously evaluated using the PDBbind v2016 core set and PDBbind v2013 core set, which serve as benchmarks, demonstrating the superior performance of ALSRI-Net. Additionally, we conducted dimensionality reduction and visualization of the extracted long-range and short-range interaction features, as well as their integrated features, providing insights into the respective contributions of these interactions and elucidating why the integration of these features leads to enhanced predictive accuracy. Our code and models are publicly available at https://github.com/Funiverse/ ALSRI-Net.

To find more 3-uncolorable vertices recognizable in polynomial time, we consider mobile 3-uncolorable vertices. In color space, the transfer is expressed by shift of dot strings in alternate configurations. An alternate configuration is the union of a DJC and movable dot strings relative to the DJC. It has several FLC one of which is a WFLC. The movement of the movable dot string exchanges FLC with WFLC to ensure that WFLC exists always in the configuration. We prove that an additional dot string (ADS) to an alternate configuration may turn the configuration to be a triplet. So, 3-uncolorable vertices in the graph of alternate configurations in planar graph with degree more than 4 G(V,E) can be recognized in O(|V8|). We need to find more recognizable 3-uncolorable vertices in polynomial time for judging 3-coloring of planar graph.

The ability of modern machine learning systems to generate contextually relevant and coherent text is often challenged when external knowledge must be integrated into the generative process. Addressing this challenge, a novel approach was developed to dynamically reduce token-level uncertainty during retrieval-augmented text generation. By modifying the Mistral model to incorporate a token-level uncertainty estimation layer, the generation process was significantly enhanced in both precision and fluency. Experiments conducted on multiple retrieval-based tasks demonstrated substantial improvements in retrieval precision, perplexity, and uncertainty reduction, showing the modified model's ability to generate more coherent outputs, particularly in scenarios involving ambiguous or incomplete retrieved information. The recalibration of token predictions during the generation process directly contributed to the model's robustness and reliability, making it more adaptable to the complexities of retrieval-augmented contexts. Results indicated that the proposed method offers a significant advancement in improving the alignment between retrieved documents and the generated responses, leading to more accurate and contextually appropriate outputs across diverse datasets.

Large Language Models (LLMs) face signicant challenges related to context length when generating code, often resulting in incoherent or incomplete outputs. This paper aims to explore the context length issue, present technical solutions, and suggest future directions for improving context retention in LLMs used for code generation.