Specific Architecture Research Articles

A Reference Architecture for Smart Car Parking Management Systems

Smart car parking management systems (SPMSs) have gained an ever-increasing popularity for the digital management of car parking processes. While various techniques and technologies have been proposed for SPMSs, the literature lacks in any generic software architecture design that can be reused systematically for the specifications of quality SPMS architectures. To bridge this gap, we propose a reference architecture (RA) for the SPMS product family after performing a comprehensive domain analysis. Our RA design offers a feature model that consists of the common and varying features for SPMSs. We offer multiple viewpoints for our RA, including context, module, component and connector, and allocation. The context viewpoint focuses on the stakeholders, the module viewpoint focuses on the software units, the component and connector viewpoint focuses on the layered architecture of SPMSs, and the allocation viewpoint focuses on mapping software units into the physical components. Each viewpoint can be re-used for specifying the application architecture of any SPMSs. We validated our RA with a real SPMS scenario specification and prototype development, where the former measures the reusability of RA and the latter measures the development performance. The RA design for SPMSs is expected to be useful for several stakeholders who research, develop, and sell SPMS solutions.

Automating skin cancer screening: a deep learning

Skin cancer presents in various forms, including squamous cell carcinoma (SCC), basal cell carcinoma (BCC), and melanoma. Established risk factors include ultraviolet (UV) radiation exposure from solar or artificial sources, lighter skin pigmentation, a history of sunburns, and a family history of the disease. Early detection and prompt intervention are crucial for achieving a favorable prognosis. Traditionally, treatment modalities include surgery, radiation therapy, and chemotherapy. Recent advancements in immunotherapy have revolutionized skin cancer diagnosis, but manual identification remains time-consuming. Artificial intelligence (AI) has shown potential in skin cancer classification, leading to automated screening methods. To support dermatologists, we improved the model for classifying images. This model is able to recognize seven different kinds of skin lesions. On the ISIC dataset, an analysis has been done. This study offers a novel approach to early skin cancer diagnosis based on image processing. Our approach leverages the high accuracy of a specific convolutional neural network architecture, utilizing transfer learning with pre-trained data to further enhance detection performance. Our findings demonstrate that the employed ResNet-50 transfer learning model achieves a remarkable accuracy of 97%, while ResNet50 without augmentation gives an accuracy of 81.57% and an F1-score of 75.75%.

The Atlas of Protein-Protein Interactions in Cancer (APPIC)-a webtool to visualize and analyze cancer subtypes.

Cancer is a complex disease with heterogeneous mutational and gene expression patterns. Subgroups of patients who share a phenotype might share a specific genetic architecture including protein-protein interactions (PPIs). We developed the Atlas of Protein-Protein Interactions in Cancer (APPIC), an interactive webtool that provides PPI subnetworks of 10 cancer types and their subtypes shared by cohorts of patients. To achieve this, we analyzed publicly available RNA sequencing data from patients and identified PPIs specific to 26 distinct cancer subtypes. APPIC compiles biological and clinical information from various databases, including the Human Protein Atlas, Hugo Gene Nomenclature Committee, g:Profiler, cBioPortal and Clue.io. The user-friendly interface allows for both 2D and 3D PPI network visualizations, enhancing the usability and interpretability of complex data. For advanced users seeking greater customization, APPIC conveniently provides all output files for further analysis and visualization on other platforms or tools. By offering comprehensive insights into PPIs and their role in cancer, APPIC aims to support the discovery of tumor subtype-specific novel targeted therapeutics and drug repurposing. APPIC is freely available at https://appic.brown.edu.

Comparative Analysis of YOLO Variants Based on Performance Evaluation for Object Detection

This study focuses on analysing and exploring the You Only Look Once (YOLO) algorithm. Specifically, this article analyses the evolution and performance of three versions (YOLOv1, YOLOv5, and YOLOv8) in object detection. The research begins by detailing the fundamental concepts of object detection and the datasets commonly used in this field. It then delves into the specific architectures and experimental outcomes associated with each YOLO version. The analysis reveals that while YOLOv8 introduces advanced features and improvements, earlier versions like YOLOv5 may offer superior stability and performance under certain conditions, particularly in specific tasks such as car detection. The discussion emphasizes the significant impact of factors such as batch size on model performance, suggesting that fine-tuning these parameters can optimize the algorithm for particular applications. The study concludes that the future of YOLO development lies in exploring and refining different variants, particularly those of YOLOv8, to better meet diverse requirements. By focusing on five distinct YOLOv8 variants, the research aims to enhance the adaptability and effectiveness of the YOLO framework across a wide range of object detection challenges, thereby contributing valuable insights into the ongoing advancement of this technology.

Gut microbial GABA imbalance emerges as a metabolic signature in mild autism spectrum disorder linked to overrepresented Escherichia.

Gut microbiota (GM) alterations have been implicated in autism spectrum disorder (ASD), yet the specific functional architecture remains elusive. Here, employing multi-omics approaches, we investigate stool samples from two distinct cohorts comprising 203 children with mild ASD or typical development. In our screening cohort, regression-based analysis for metabolomic profiling identifies an elevated γ-aminobutyric acid (GABA) to glutamate (Glu) ratio as a metabolic signature of ASD, independent of age and gender. In the validating cohort, we affirm the GABA/Glu ratio as an ASD diagnostic indicator after adjusting for geography, age, gender, and specific food-consuming frequency. Integrated analysis of metabolomics, 16S rRNA sequencing, and metagenomics reveals a correlation between overrepresented Escherichia and disrupted GABA metabolism. Furthermore, we observe social behavioral impairments in weaning mice transplanted with E.coli, suggesting a potential link to ASD symptomatology. Collectively, these findings provide insights into potential diagnostic and therapeutic strategies aimed at evaluating and restoring gut microbial neurotransmitter homeostasis.

Visualizing the DNA Damage Response in Purkinje Cells Using Cerebellar Organotypic Cultures.

Cerebellar Purkinje cells (PCs) exhibit a unique interplay of high metabolic rates, specific chromatin architecture, and extensive transcriptional activity, making them particularly vulnerable to DNA damage. This necessitates an efficient DNA damage response (DDR) to prevent cerebellar degeneration, often initiated by PC dysfunction or loss. A notable example is the genome instability syndrome, ataxia-telangiectasia (A-T), marked by progressive PC depletion and cerebellar deterioration. Investigating DDR mechanisms in PCs is vital for elucidating the pathways leading to their degeneration in such disorders. However, the complexity of isolating and cultivating PCs in vitro has long hindered research efforts. Murine cerebellar organotypic (slice) cultures offer a feasible alternative, closely mimicking the in vivo tissue environment. Yet, this model is constrained to DDR indicators amenable to microscopic imaging. We have refined the organotypic culture protocol, demonstrating that fluorescent imaging of protein-bound poly(ADP-ribose) (PAR) chains, a rapid and early DDR indicator, effectively reveals DDR dynamics in PCs within these cultures, in response to genotoxic stress.

The Effect of Processing Techniques on the Classification Accuracy of Brain-Computer Interface Systems.

Background/Objectives: Accurately classifying Electroencephalography (EEG) signals is essential for the effective operation of Brain-Computer Interfaces (BCI), which is needed for reliable neurorehabilitation applications. However, many factors in the processing pipeline can influence classification performance. The objective of this study is to assess the effects of different processing steps on classification accuracy in EEG-based BCI systems. Methods: This study explores the impact of various processing techniques and stages, including the FASTER algorithm for artifact rejection (AR), frequency filtering, transfer learning, and cropped training. The Physionet dataset, consisting of four motor imagery classes, was used as input due to its relatively large number of subjects. The raw EEG was tested with EEGNet and Shallow ConvNet. To examine the impact of adding a spatial dimension to the input data, we also used the Multi-branch Conv3D Net and developed two new models, Conv2D Net and Conv3D Net. Results: Our analysis showed that classification accuracy can be affected by many factors at every stage. Applying the AR method, for instance, can either enhance or degrade classification performance, depending on the subject and the specific network architecture. Transfer learning was effective in improving the performance of all networks for both raw and artifact-rejected data. However, the improvement in classification accuracy for artifact-rejected data was less pronounced compared to unfiltered data, resulting in reduced precision. For instance, the best classifier achieved 46.1% accuracy on unfiltered data, which increased to 63.5% with transfer learning. In the filtered case, accuracy rose from 45.5% to only 55.9% when transfer learning was applied. An unexpected outcome regarding frequency filtering was observed: networks demonstrated better classification performance when focusing on lower-frequency components. Higher frequency ranges were more discriminative for EEGNet and Shallow ConvNet, but only when cropped training was applied. Conclusions: The findings of this study highlight the complex interaction between processing techniques and neural network performance, emphasizing the necessity for customized processing approaches tailored to specific subjects and network architectures.

Learning Generative Models Using Denoising Density Estimators.

Learning probabilistic models that can estimate the density of a given set of samples, and generate samples from that density, is one of the fundamental challenges in unsupervised machine learning. We introduce a new generative model based on denoising density estimators (DDEs), which are scalar functions parametrized by neural networks, that are efficiently trained to represent kernel density estimators of the data. Leveraging DDEs, our main contribution is a novel technique to obtain generative models by minimizing the Kullback-Leibler (KL)-divergence directly. We prove that our algorithm for obtaining generative models is guaranteed to converge consistently to the correct solution. Our approach does not require specific network architecture as in normalizing flows (NFs), nor use ordinary differential equation (ODE) solvers as in continuous NFs. Experimental results demonstrate substantial improvement in density estimation and competitive performance in generative model training.

Gene mobility elements mediate cell type specific genome organization and radial gene movement in vivo.

Understanding the level of genome organization that governs gene regulation remains a challenge despite advancements in chromatin profiling techniques. Cell type specific chromatin architectures may be obscured by averaging heterogeneous cell populations. Here we took a reductionist perspective, starting with the relocation of the hunchback gene to the nuclear lamina in Drosophila neuroblasts. We previously found that this event terminates competence to produce early-born neurons and is mediated by an intronic 250 base-pair element, which we term gene mobility element (GME). Here we found over 800 putative GMEs globally that are chromatin accessible and are Polycomb (PcG) target sites. GMEs appear to be distinct from PcG response elements, however, which are largely chromatin inaccessible in neuroblasts. Performing in situ Hi-C of purified neuroblasts, we found that GMEs form megabase-scale chromatin interactions, spanning multiple topologically associated domain borders, preferentially contacting other GMEs. These interactions are cell type and stage-specific. Notably, GMEs undergo developmentally- timed mobilization to/from the neuroblast nuclear lamina, and domain swapping a GFP reporter transgene intron with a GME relocates the transgene to the nuclear lamina in embryos. We propose that GMEs constitute a genome organizational framework and mediate gene-to-lamina mobilization during progenitor competence state transitions in vivo .

An implemented predictive coding model of lexico-semantic processing explains the dynamics of univariate and multivariate activity within the left ventromedial temporal lobe during reading comprehension.

During language comprehension, the larger neural response to unexpected versus expected inputs is often taken as evidence for predictive coding-a specific computational architecture and optimization algorithm proposed to approximate probabilistic inference in the brain. However, other predictive processing frameworks can also account for this effect, leaving the unique claims of predictive coding untested. In this study, we used MEG to examine both univariate and multivariate neural activity in response to expected and unexpected inputs during word-by-word reading comprehension. We further simulated this activity using an implemented predictive coding model that infers the meaning of words from their orthographic form. Consistent with previous findings, the univariate analysis showed that, between 300-500ms, unexpected words produced a larger evoked response than expected words within a left ventromedial temporal region that supports the mapping of orthographic word-forms onto lexical and conceptual representations. Our model explained this larger evoked response as the enhanced lexico-semantic prediction error produced when prior top-down predictions failed to suppress activity within lexical and semantic "error units". Critically, our simulations showed that, despite producing minimal prediction error, expected inputs nonetheless reinstated top-down predictions within the model's lexical and semantic "state" units. Two types of multivariate analyses provided evidence for this functional distinction between state and error units within the ventromedial temporal region. First, within each trial, individual voxels within this region produced unique temporal patterns between 300-500ms that that resembled the temporal patterns produced within a pre-activation time-window before the expected input became available. First, within each trial, the same left ventromedial voxels that produced a larger response between 300-500ms to unexpected inputs, reinstated unique pre-activated item-specific temporal patterns in response to expected inputs within this same time window. Second, across trials, and again within the same 300-500ms time window and left ventromedial temporal region, pairs of expected words produced spatial patterns that were more similar to one another than the spatial patterns produced by pairs of expected and unexpected words, regardless of specific item. Together, these findings provide compelling evidence that the left ventromedial temporal lobe employs predictive coding to infer the meaning of incoming words from their orthographic form during reading comprehension.

Bilinear Models of Parts and Appearances in Generative Adversarial Networks.

Recent advances in the understanding of Generative Adversarial Networks (GANs) have led to remarkable progress in visual editing and synthesis tasks, capitalizing on the rich semantics that are embedded in the latent spaces of pre-trained GANs. However, existing methods are often tailored to specific GAN architectures and are limited to either discovering global semantic directions that do not facilitate localized control, or require some form of supervision through manually provided regions or segmentation masks. In this light, we present an architecture-agnostic approach that jointly discovers factors representing spatial parts and their appearances in an entirely unsupervised fashion. These factors are obtained by applying a semi-nonnegative tensor factorization on the feature maps, which in turn enables context-aware local image editing with pixel-level control. In addition, we show that the discovered appearance factors correspond to saliency maps that localize concepts of interest, without using any labels. Experiments on a wide range of GAN architectures and datasets show that, in comparison to the state of the art, our method is far more efficient in terms of training time and, most importantly, provides much more accurate localized control.

Deep Learning-based Gold Price Prediction: A Novel Approach using Time Series Analysis

This paper presents a deep learning-based system for predicting gold prices using historical data. The system leverages Long Short-Term Memory (LSTM), a specialized recurrent neural network architecture, to capture temporal dependencies and patterns in the time series data of gold prices. A comprehensive dataset of historical gold prices is used, and the model is trained on a sequence of past data points to predict future prices. The data is preprocessed using normalization techniques to improve the performance of the model. Experimental results demonstrate the effectiveness of the proposed model in providing accurate price predictions, offering potential utility in financial forecasting and decision-making processes. The system's performance is evaluated through visualization and statistical metrics, illustrating its capacity to track gold price trends and predict future market movements. This work contributes to the growing field of time series forecasting by applying deep learning techniques to financial markets.

Low‐frequency noise and deep‐level transient spectroscopy in LWIR Auger‐suppressed Hg1-xCdxTe heterostructure detector

Low-frequency noise spectroscopy (LFNS) along with deep-level transient spectroscopy (DLTS) are complementary and effective tools to study and characterize the carrier traps in semiconductors. These traps caused, e.g., by contamination by foreign atoms or various types of dislocations, can significantly affect quantum efficiency, dark current, responsivity, and noise generated by devices especially when operating under bias. Since DLTS is difficult to apply in high leakage current devices, LFNS can be used to overcome this limitation, so the use of both methods gives very effective and reliable results during research on various devices. In this paper, we reported a study of defects activation energies in HgCdTe Auger-suppressed long-wavelength infrared (LWIR) heterostructure-based detector using these two experimental methods. By proper structure design, the examined detector was optimized for high operating temperature (HOT) conditions ≥ 200 K. The results obtained showed that in such detectors, grown by the metal organic chemical vapor deposition (MOCVD) technique, a few traps can be extracted. The found trap levels and activation energies were located below and above the absorber bandgap, so they can be identified in both absorber and other heterostructure layers. Due to specific multilayer architecture, a precise interpretation of the results is difficult. Nevertheless, the most probable trap locations based on the current state of knowledge were discussed and proposed.

Unconventional C-Hlg···H-C (Hlg = Cl, Br, and I) Interactions Involving Organic Halides: A Theoretical Study.

In this study, unconventional C-Hlg···H-C (Hlg = Cl, Br, and I) interactions involving sp, sp2, and sp3 organic halides were investigated at the RI-MP2/aug-cc-pVQZ level of theory. Energy Decomposition Analyses (EDA) and Natural Bonding Orbital (NBO) studies showed that these intermolecular contacts are mainly supported by orbital and dispersion contributions, which counteracted the unfavorable/slightly favorable electrostatics due to the halogen-hydrogen σ-hole facing. In addition, the Bader's Quantum Theory of Atoms in Molecules (QTAIM) and the Noncovalent Interaction plot (NCIplot) visual index methodologies were used to further characterize the interactions discussed herein. We expect that the results reported herein will be useful in the fields of supramolecular chemistry, crystal engineering, and rational drug design, where the fine tuning of noncovalent interactions is crucial to achieve molecular recognition or a specific solid-state architecture.

Enhancing satellite networks with deep reinforcement learning: A focus on IoT connectivity and dynamic resource management

The integration of satellite-based networks into our communication infrastructure is increasingly becoming standard practice, driven by the widespread adoption of 5G technologies and the escalating demand for continuous content delivery. Achieving this integration efficiently necessitates significant improvements in reducing latency and enhancing data throughput. In response to these challenges, we propose the development of a specialized satellite architecture alongside a novel algorithm focused on call access and admission control within Low Earth Orbit (LEO) satellite systems. This solution leverages an Artificial Intelligence (AI) driven approach, employing a Deep Reinforcement Learning (DRL) agent to automate satellite operations through advanced beam localization techniques. Furthermore, we introduce a beam protocol that seamlessly integrates this automated learning mechanism into its operational framework. The efficacy of our proposed algorithm is rigorously evaluated through simulation studies of a multibeam satellite system, demonstrating the potential of Deep Reinforcement Learning in facilitating dynamic resource allocation with improved efficiency.

The Antecedents of Transformer Models

Transformer models of language represent a step change in our ability to account for cognitive phenomena. Although the specific architecture that has garnered recent interest is quite young, many of its components have antecedents in the cognitive science literature. In this article, we start by providing an introduction to large language models aimed at a general psychological audience. We then highlight some of the antecedents, including the importance of scale, instance-based memory models, paradigmatic association and systematicity, positional encodings of serial order, and the learning of control processes. This article offers an exploration of the relationship between transformer models and their precursors, showing how they can be understood as a next phase in our understanding of cognitive processes.

Facile Synthesis of Polypyrrole-Decorated RGO-CuS Nanocomposite for Efficient Nickel Removal from Wastewater.

Efficient wastewater treatment, particularly the removal of heavy metal ions, remains a challenging priority in environmental remediation. This study introduces a novel sandwich-structured nanocomposite, RGO-CuS-PPy, composed of reduced graphene oxide (RGO), copper sulfide (CuS), and polypyrrole (PPy), synthesized via a straightforward hydrothermal method. The unique combination of RGO, CuS, and PPy offers enhanced adsorption capacity for Ni(II) ions due to RGO's high surface area and CuS's active binding sites, supported by PPy's structural stability contributions. This study is among the first to explore this specific nanocomposite architecture for Ni(II) removal, achieving an adsorption capacity of 166.67 mg/g and a high removal efficiency of 94.9% within 210 min for 55 mg/L of Ni(II) concentration at pH 6 and adsorbent dose of 3 mg/15 mL. The kinetic analysis shows the best fitted time-dependent experimental data with the pseudo-second-order model, indicating chemisorption. Isotherm studies confirmed the Langmuir model as the best fit, yielding a high monolayer adsorption capacity of 166.67 mg/g. Thermodynamic analysis shows the adsorption process was endothermic (ΔH° = 80.23 kJ/mol) and spontaneous (ΔG° ranging from -6.985 to -14.399 kJ/mol). Additionally, reusability tests using 0.1 M HCl for desorption demonstrated good reusability, emphasizing the RGO-CuS-PPy nanocomposite's potential as a sustainable adsorbent for Ni(II) removal in wastewater treatment applications.

Distinguishing between children referred for assessment of excessive daytime sleepiness using polysomnographic measures

BackgroundCurrently, diagnosing narcolepsy and idiopathic hypersomnia (IH) in subjectively sleepy children requires an overnight polysomnographic sleep study followed by a daytime multiple sleep latency test (MSLT). We aimed to compare sleep macro-architecture to identify differences between these groups. MethodsAll children referred for a MSLT between May 2010 to December 2023 whose parent consented for their data be used in research were eligible. Each child was age- and sex-matched to a control. Sleep stability was defined as the maintenance of a particular sleep stage before waking or transitioning to another sleep state. As a measure of sleep disturbance, the number and duration of bouts of each sleep stage was recorded. Results28 children with Narcolepsy, 11 with IH and 26 with subjective sleepiness were included. Children with narcolepsy exhibited higher numbers of transitions to wake after sleep onset compared to their controls and to the subjectively sleepy group (p < 0.001 for both). The number of REM bouts was greater in the narcolepsy group compared to their control group (p < 0.001), the IH group (p < 0.05) and the subjectively sleepy group (p < 0.05), while the average duration of REM bouts was shorter in the narcolepsy group compared to both the IH group (p < 0.05) and subjectively sleepy group (p < 0.05). Mean sleep latency on the MSLT was correlated with a number of polysomnographic variables. ConclusionsOur study suggests that specific sleep architecture patterns could potentially serve as diagnostic biomarkers for distinguishing paediatric narcolepsy from those with IH and those with a non-diagnostic MSLT.

Approximating families of sharp solutions to Fisher's equation with physics-informed neural networks

This paper employs physics-informed neural networks (PINNs) to solve Fisher's equation, a fundamental reaction-diffusion system with both simplicity and significance. The focus is on investigating Fisher's equation under conditions of large reaction rate coefficients, where solutions exhibit steep traveling waves that often present challenges for traditional numerical methods. To address these challenges, a residual weighting scheme is introduced in the network training to mitigate the difficulties associated with standard PINN approaches. Additionally, a specialized network architecture designed to capture traveling wave solutions is explored. The paper also assesses the ability of PINNs to approximate a family of solutions by generalizing across multiple reaction rate coefficients. The proposed method demonstrates high effectiveness in solving Fisher's equation with large reaction rate coefficients and shows promise for meshfree solutions of generalized reaction-diffusion systems.

Nanoarchitecture via Synchronic Stacking of Metallic and Nonmetallic Two-Dimensional Nanosheets for Optimal Light-Driven Ion Transport.

The exceptional selectivity and responsive ion transport in biological channels inspire technology breakthrough in energy, environmental, and resource sectors. However, existing nanofluidic systems with a high photothermal conversion efficiency often exhibit excessive thermal conductivity, which impedes the creation of effective temperature gradients and results in a low ion transport efficiency. In this study, a strategy based on the synchronic stacking of metallic and nonmetallic two-dimensional (2D) nanosheets was presented to construct heterogeneous nanofluidic channels. This specific nanoconfined architecture sustained high temperatures in the illuminated area while maintaining low temperatures in the nonilluminated area, thus obtaining a robust driving force from sunlight for directional ion transport. As a result, our light-responsive ion transport system demonstrated significant potential in solar energy conversion and osmotic energy harvesting, surpassing those of all previously reported nanofluidic systems. Additionally, although it is still at the proof-of-concept stage, it shows great promise in light signal monitoring. This work provides an effective strategy for developing advanced light-responsive ion transport systems and their important applications.