Small Database Research Articles

An efficient edge computing management mechanism for sustainable smart cities

Since ancient times, humanity has envisioned a world where people and things connect and interact. The advent of generation mobile systems, the so-called 5 G in the early century, has realized this dream. 5 G can provide network services with extremely-high throughput and extremely-low delay and allows a huge device number to connect based on Internet infrastructure, forming the Internet of Things (IoT). In recent years, IoT has been applied in various fields serving humanity, such as smart cities, smart agriculture, healthcare, education, military, and IoT ecosystems. One of the main challenges of IoT applications is computing solutions to reduce service response times. This study proposes an efficient edge computing management mechanism for IoT applications in smart cities. Our mechanism proposes a small database (called an information map) that allows edge computing servers of smart cities to store edge service information. When the mobile end-users move to the new edge server' managed coverage, properties related to the EC service are exchanged between the edge servers. The experiment results have shown that our proposed mechanism improves service response time and energy consumption many times compared to the traditional mechanism. We hope that this mechanism will be strongly applied to sustainable smart cities in the future.

Jet Flow and Noise Predictions for the Doak Laboratory Experiment

Large-eddy simulations (LESs) are performed for two isolated unheated jet flows corresponding to a Doak Laboratory experiment performed at the University of Southampton. The jet speeds studied correspond to acoustic Mach numbers of 0.6 and 0.8 as well as Reynolds numbers based on the nozzle exit diameter of about one million. The LES method is based on the compact accurately boundary-adjusting high-resolution technique (CABARET) and is implemented on graphics processing units (GPUs) to obtain 1000–1100 convective time units for statistical averaging with reasonable run times. In comparison with the previous jet LES calculations with the GPU CABARET method, the mean-flow velocity and turbulent intensity profiles are matched with the hot-wire measurements just downstream of the nozzle exit. The far-field noise spectra of the Doak jets are evaluated using two methods: the Ffowcs Williams–Hawkings approach and a reduced-order implementation of the Goldstein generalized acoustic analogy. The flow and noise results are compared with hot-wire and acoustic microphone measurements of the Doak Laboratory and critically analyzed in comparison with the NASA small hot jet acoustic rig database.

Machine learning and reduced order modelling for the simulation of braided stent deployment.

Endoluminal reconstruction using flow diverters represents a novel paradigm for the minimally invasive treatment of intracranial aneurysms. The configuration assumed by these very dense braided stents once deployed within the parent vessel is not easily predictable and medical volumetric images alone may be insufficient to plan the treatment satisfactorily. Therefore, here we propose a fast and accurate machine learning and reduced order modelling framework, based on finite element simulations, to assist practitioners in the planning and interventional stages. It consists of a first classification step to determine a priori whether a simulation will be successful (good conformity between stent and vessel) or not from a clinical perspective, followed by a regression step that provides an approximated solution of the deployed stent configuration. The latter is achieved using a non-intrusive reduced order modelling scheme that combines the proper orthogonal decomposition algorithm and Gaussian process regression. The workflow was validated on an idealized intracranial artery with a saccular aneurysm and the effect of six geometrical and surgical parameters on the outcome of stent deployment was studied. We trained six machine learning models on a dataset of varying size and obtained classifiers with up to 95% accuracy in predicting the deployment outcome. The support vector machine model outperformed the others when considering a small dataset of 50 training cases, with an accuracy of 93% and a specificity of 97%. On the other hand, real-time predictions of the stent deployed configuration were achieved with an average validation error between predicted and high-fidelity results never greater than the spatial resolution of 3D rotational angiography, the imaging technique with the best spatial resolution (0.15mm). Such accurate predictions can be reached even with a small database of 47 simulations: by increasing the training simulations to 147, the average prediction error is reduced to 0.07mm. These results are promising as they demonstrate the ability of these techniques to achieve simulations within a few milliseconds while retaining the mechanical realism and predictability of the stent deployed configuration.

Synthesis of a Small Fingerprint Database through a Deep Generative Model for Indoor Localisation

In deep learning (DL), the deep generative model is helpful for data augmentation objectives to tackle the lack of datasets that have a significant impact on learning performance. Data augmentation or synthesis is expected to solve the issue in a small/sparse database. The problem of databasing also exists in the fingerprint-based indoor localisation system. The dense offline fingerprint database must be constructed with the accuracy requirement. However, this will affect the high cost, massive laborious work, and increase the complexity of the system. Therefore, this paper proposes to address these issues by generating synthetic data via a deep generative model. The generative adversarial network (GAN) is selected to generate the synthetic fingerprint database for indoor localisation. Our database consideration consists of power-based parameters, i.e., the received signal strength indicator (RSSI) from Wi-Fi devices obtained from the actual measurement campaign. Some of the literature mainly discusses how GAN works in a vast and complex dataset. Here, we consider applying GAN in a relatively small dataset and for a simple setup. Our results show that by only using the 20 % fraction of actual RSSI data combined with the synthetic RSSI, the accuracy validation performance is slightly higher than when using all actual data usage. Moreover, in only 60 % of actual data usage and in combination with 625 samples of synthetic data, the accuracy performance is improved to 0.73 (1.37 times higher than the use of all actual data, 0.53). Thus, this result proves that the challenges of offline fingerprint databases can be alleviated by data synthesis through GAN by using only a small dataset.

COVID-19 therapy optimization by AI-driven biomechanical simulations.

The COVID-19 disease causes pneumonia in many patients that in the most serious cases evolves into the Acute Distress Respiratory Syndrome (ARDS), requiring assisted ventilation and intensive care. In this context, identification of patients at high risk of developing ARDS is a key point for early clinical management, better clinical outcome and optimization in using the limited resources available in the intensive care units. We propose an AI-based prognostic system that makes predictions of oxygen exchange with arterial blood by using as input lung Computed Tomography (CT), the air flux in lungs obtained from biomechanical simulations and Arterial Blood Gas (ABG) analysis. We developed and investigated the feasibility of this system on a small clinical database of proven COVID-19 cases where the initial CT and various ABG reports were available for each patient. We studied the time evolution of the ABG parameters and found correlation with the morphological information extracted from CT scans and disease outcome. Promising results of a preliminary version of the prognostic algorithm are presented. The ability to predict the evolution of patients' respiratory efficiency would be of crucial importance for disease management.

Physiologically-Based Kinetic Modeling of Intravenously Administered Gold (Au) Nanoparticles.

Physiologically-based kinetic (PBK) modeling is a valuable tool to understand the kinetics of nanoparticles (NPs) in vivo. However, estimating PBK parameters remains challenging and commonly requires animal studies. To develop predictive models to estimate PBK parameter values based on NP characteristics, a database containing PBK parameter values and corresponding NP characteristics is needed. As a first step toward this objective, this study estimates PBK parameters for gold NPs (AuNPs) and provides a comparison of two different NPs. Two animal experiments are conducted in which varying doses of AuNPs attached with polyethylene glycol (PEG) are administered intravenously to rats. The resulting Au concentrations are used to estimate PBK model parameters. The parameters are compared with PBK parameters previously estimated for poly(alkyl cyanoacrylate) NPs loaded with cabazitaxel and for LipImage 815. This study shows that a small initial database of PBK parameters collected for three NPs is already sufficient to formulate new hypotheses on NP characteristics that may be predictive of PBK parameter values. Further research should focus on developing a larger database and on developing quantitative models to predict PBK parameter values.

Genomic prediction in Brazilian Guzerá cattle: application of a single-step approach to productive and reproductive traits.

This study aimed to investigate the feasibility of genomic prediction for productive and reproductive traits in Guzerá cattle using single-step genomic best linear unbiased prediction (ssGBLUP). Evaluations included the 305-day cumulative yields (first lactation, in kg) of milk, lactose, protein, fat, and total solids; adjusted body weight (kg) at the ages of 450, 365, and 210days; and age at first calving (in days), from a database containing 197,283 measurements from Guzerá males and females born between 1954 and 2018. The pedigree included 433,823 animals spanning up to 14 overlapping generations. A total of 1618 animals were genotyped. The analyses were performed using ssGBLUP and traditional BLUP methods. Predictive ability and bias were accessed using cross-validation: predictive ability was similar between the methods and ranged from 0.27 to 0.47 for the genomic-based model and from 0.30 to 0.45 for the pedigree-based model; the bias was also similar between the methods, ranging from 0.88 to 1.35 in the genomic-based model and from 0.96 to 1.41 in the pedigree-based model. The individual accuracies of breeding values were evidently increased in the genomic evaluation, with values ranging from 0.41 to 0.56 in the genomic-based model and from 0.26 to 0.54 in the pedigree-based model. Even based on a small number of genotyped animals and a small database for some traits, the results suggest that ssGBLUP is feasible and may be applied to national genetic evaluation of the breed to increase the accuracy of breeding values without greatly impacting predictive ability and bias.

Discovery of novel A2AR antagonists through deep learning-based virtual screening

The A2A adenosine receptor (A2AR) is emerging as a promising drug target for cancer immunotherapy. Novel A2AR antagonists are highly demanded due to few candidates entering clinic trials specific for cancer treatment. Structure-based virtual screening has made a great contribution to discover novel A2AR antagonists, but most depended on inefficient molecular docking on relatively small molecular databases. In this work, a deep learning strategy was applied to accelerate docking-based virtual screening, through which new structural types of A2AR antagonists for an extremely large molecular library were found successfully.

Identification of NS2B-NS3 Protease Inhibitors for Therapeutic Application in ZIKV Infection: A Pharmacophore-Based High-Throughput Virtual Screening and MD Simulations Approaches

Zika virus (ZIKV) pandemic and its implication in congenital malformations and severe neurological disorders had created serious threats to global health. ZIKV is a mosquito-borne flavivirus which spread rapidly and infect a large number of people in a shorter time-span. Due to the lack of effective therapeutics, this had become paramount urgency to discover effective drug molecules to encounter the viral infection. Various anti-ZIKV drug discovery efforts during the past several years had been unsuccessful to develop an effective cure. The NS2B-NS3 protein was reported as an attractive therapeutic target for inhibiting viral proliferation, due to its central role in viral replication and maturation of non-structural viral proteins. Therefore, the current in silico drug exploration aimed to identify the novel inhibitors of Zika NS2B-NS3 protease by implementing an e-pharmacophore-based high-throughput virtual screening. A 3D e-pharmacophore model was generated based on the five-featured (ADPRR) pharmacophore hypothesis. Subsequently, the predicted model is further subjected to the high-throughput virtual screening to reveal top hit molecules from the various small molecule databases. Initial hits were examined in terms of binding free energies and ADME properties to identify the candidate hit exhibiting a favourable pharmacokinetic profile. Eventually, molecular dynamic (MD) simulations studies were conducted to evaluate the binding stability of the hit molecule inside the receptor cavity. The findings of the in silico analysis manifested affirmative evidence for three hit molecules with -64.28, -55.15 and -50.16 kcal/mol binding free energies, as potent inhibitors of Zika NS2B-NS3 protease. Hence, these molecules holds the promising potential to serve as a prospective candidates to design effective drugs against ZIKV and related viral infections.

Small Tool Image Database and Object Detection Approach for Indoor Construction Site Safety

Object detection refers to computer vision technology that identifies the class and location of objects in an image. With the development of deep-learning techniques, object detection has been widely studied in the construction industry and has mainly focused on detecting heavy equipment and workers at outdoor sites. Therefore, an object detection model for small tools in indoor sites, where various accidents frequently occur owing to them, is required to enhance safety. In particular, many high-quality image databases are important for developing object detection approaches for small tools. In this study, 12 small tools commonly used in indoor construction were selected, and a database of 34,738 images of the selected tools was created and used to train the object detection models. YOLOv5x and YOLOv5s are selected as object detection models, for which the mean average precision with the 0.5 intersection-over-union value were 69.1% and 64.4%, respectively, and their frame rates were 58.82 and 142.86 frame per second, respectively. The results demonstrate that the established database is appropriate for the development of object detection for small tools and that object detection can be used for real-time safety management of construction indoor sites.

Next generation drug connectivity mapping for acquiring therapeutic agents to differentially regulate myelination.

Next generation drug connectivity mapping for acquiring therapeutic agents to differentially regulate myelination.

Правый берег Исети в районе общественной территории «сад Нурова» у Царского моста в городе Екатеринбурге: история городской застройки и научных исследований 1770–2023 гг.

The Unified State Register of Cultural Heritage Objects in Ekaterinburg includes about 800 sites. In recent years, the public has become more likely to talk about the need to preserve cultural heritage, because many of these objects are in poor condition. Despite the public outcry, various parks and green areas located in historical buildings are under particular threat of extinction. One of the main problems when conducting research on such territories is a small database of archival sources, which may complicate the task of substantiating the historical value of the object. Such territories include a section of the right bank of the Iset River near the Tsarsky Bridge within the Dekabristov — Tveritin Streets of Ekaterinburg, now also known as the Nurov Garden public territory. Formed in the middle of the 18th century as commercial quarters, this space transformed into a string of private enclosed gardens of the city’s influential merchants with a rich ensemble of buildings: gazebos and pavilions, baths and berths for boats, bridges and piers. Now, having left behind the time of complete oblivion, preparations for the process of landscaping the embankment begins on the territory. The article, relying primarily on documentary archival sources, considers the history of the emergence and development of this territory, analyzes the affecting factors and summarizes the studies already conducted in this territory.

An Efficient Few-Shot Object Detection Method for Railway Intrusion via Fine-Tune Approach and Contrastive Learning

Railway intrusion detection plays an important role in the railway intelligent transportation system, assisting the safe operation of trains. The existing deep learning-based object detection methods excessively rely on a large number of supervision samples, and build up such a railway intrusion dataset is usually costly. The emerging few-shot object detection (FSOD) can realize object detection task by learning a small number of samples, which is a suitable candidate for railway intrusion detection. However, small databases lead to poor classification performances, and most existing FSOD methods have the drawbacks of high computational complexity and slow adaptation speed. To deal with these issues, an efficient few-shot object detection method for railway intrusion (FSRD) is proposed in this paper, extending FSOD to railway intrusion detection with few training samples to achieve good performance. A simple yet effective fine-tuning method is also developed to improve overall performance. There is a classifier head initializer to perform an appropriate initialization of the model weights and speed up the adaptation of the detector. The contrastive learning method is further introduced to cope with the misclassification. The proposed novel method is tested on a real railway intrusion dataset. Experimental results show that it achieves an average detection accuracy of 80.51 mAP50 under the setting of 30 shots. Compared with the state-of-the-art (SOTA) FSOD methods, it is capable of yielding the optimal and suboptimal performance in all settings. Furthermore, conducting extensive experiments demonstrates its good detection accuracy, fast adaptability, and satisfactory generalization ability in railway intrusion detection.

Preparing a database of corrected protein structures important in cell signaling pathways.

Precise structures of macromolecules are important for structure-based drug design. Due to the limited resolution of some structures obtained from X-ray diffraction crystallography, differentiation between the NH and O atoms can be difficult. Sometimes a number of amino acids are missing from the protein structure. In this research, we intend to introduce a small database that we have prepared for providing the corrected 3D structure files of proteins frequently used in structure-based drug design protocols. 3454 soluble proteins belonging to the cancer signaling pathways were collected from the PDB database from which a dataset of 1001 was obtained. All were subjected to corrections in the protein preparation step. 896 protein structures out of 1001 were corrected successfully and the decision on the remained 105 proposed twelve for homology modeling to correct the missing residues. Three of them were subjected to molecular dynamics simulation for 30 ns. 896 corrected proteins were perfect and homology modeling on 12 proteins with missing residues in the backbone resulted in acceptable models according to Ramachandran, z-score, and DOPE energy plots. RMSD, RMSF, and Rg values verified the stability of the models after 30 ns molecular dynamics simulation. A collection of 1001 proteins were modified for some defects such as adjustment of the bond orders and formal charges, and addition of missing side chains of residues. Homology modeling corrected the amino missing backbone residues. This database will be completed for quite a lot of water-soluble proteins to be uploaded to the internet.

Computer-assisted multistep chemoenzymatic retrosynthesis using a chemical synthesis planner.

Chemoenzymatic synthesis methods use organic and enzyme chemistry to synthesize a desired small molecule. Complementing organic synthesis with enzyme-catalyzed selective transformations under mild conditions enables more sustainable and synthetically efficient chemical manufacturing. Here, we present a multistep retrosynthesis search algorithm to facilitate chemoenzymatic synthesis of pharmaceutical compounds, specialty chemicals, commodity chemicals, and monomers. First, we employ the synthesis planner ASKCOS to plan multistep syntheses starting from commercially available materials. Then, we identify transformations that can be catalyzed by enzymes using a small database of biocatalytic reaction rules previously curated for RetroBioCat, a computer-aided synthesis planning tool for biocatalytic cascades. Enzymatic suggestions captured by the approach include ones capable of reducing the number of synthetic steps. We successfully plan chemoenzymatic routes for active pharmaceutical ingredients or their intermediates (e.g., Sitagliptin, Rivastigmine, and Ephedrine), commodity chemicals (e.g., acrylamide and glycolic acid), and specialty chemicals (e.g., S-Metalochlor and Vanillin), in a retrospective fashion. In addition to recovering published routes, the algorithm proposes many sensible alternative pathways. Our approach provides a chemoenzymatic synthesis planning strategy by identifying synthetic transformations that could be candidates for enzyme catalysis.

Refocus attention span networks for handwriting line recognition

Recurrent neural networks have achieved outstanding recognition performance for handwriting identification despite the enormous variety observed across diverse handwriting structures and poor-quality scanned documents. We initially proposed a BiLSTM baseline model with a sequential architecture well-suited for modeling text lines due to its ability to learn probability distributions over character or word sequences. However, employing such recurrent paradigms prevents parallelization and suffers from vanishing gradients for long sequences during training. To alleviate these limitations, we propose four significant contributions to this work. First, we devised an end-to-end model composed of a split-attention CNN-backbone that serves as a feature extraction method and a self-attention Transformer encoder–decoder that serves as a transcriber method to recognize handwriting manuscripts. The multi-head self-attention layers in an encoder–decoder transformer-based enhance the model’s ability to tackle handwriting recognition and learn the linguistic dependencies of character sequences. Second, we conduct various studies on transfer learning (TL) from large datasets to a small database, determining which model layers require fine-tuning. Third, we attained an efficient paradigm by combining different strategies of TL with data augmentation (DA). Finally, since the robustness of the proposed model is lexicon-free and can recognize sentences not presented in the training phase, the model is only trained on a few labeled examples with no extra cost of generating and training on synthetic datasets. We recorded comparable and outperformed Character and Word Error Rates CER/WER on four benchmark datasets to the most recent (SOTA) models.

A few-shot approach for COVID-19 screening in standard and portable chest X-ray images

Reliable and effective diagnostic systems are of vital importance for COVID-19, specifically for triage and screening procedures. In this work, a fully automatic diagnostic system based on chest X-ray images (CXR) has been proposed. It relies on the few-shot paradigm, which allows to work with small databases. Furthermore, three components have been added to improve the diagnosis performance: (1) a region proposal network which makes the system focus on the lungs; (2) a novel cost function which adds expert knowledge by giving specific penalties to each misdiagnosis; and (3) an ensembling procedure integrating multiple image comparisons to produce more reliable diagnoses. Moreover, the COVID-SC dataset has been introduced, comprising almost 1100 AnteroPosterior CXR images, namely 439 negative and 653 positive according to the RT-PCR test. Expert radiologists divided the negative images into three categories (normal lungs, COVID-related diseases, and other diseases) and the positive images into four severity levels. This entails the most complete COVID-19 dataset in terms of patient diversity. The proposed system has been compared with state-of-the-art methods in the COVIDGR-1.0 public database, achieving the highest accuracy (81.13% ± 2.76%) and the most robust results. An ablation study proved that each system component contributes to improve the overall performance. The procedure has also been validated on the COVID-SC dataset under different scenarios, with accuracies ranging from 70.81 to 87.40%. In conclusion, our proposal provides a good accuracy appropriate for the early detection of COVID-19.

Joint coupled representation and homogeneous reconstruction for multi-resolution small sample face recognition

Off-the-shelf dictionary learning algorithms have achieved satisfactory results in small sample face recognition applications. However, the achieved results depend on the facial images obtained at a single resolution. In practice, the resolution of the images captured on the same target is different because of the different shooting equipment and different shooting distances. These images of the same category at different resolutions will pose a great challenge to these algorithms. In this paper, we propose a Joint Coupled Representation and Homogeneous Reconstruction (JCRHR) for multi-resolution small sample face recognition. In JCRHR, an analysis dictionary is introduced and combined with the synthetic dictionary for coupled representation learning, which better reveals the relationship between coding coefficients and samples. In addition, a coherence enhancement term is proposed to improve the coherent representation of the coding coefficients at different resolutions, which facilitates the reconstruction of the sample by its homogeneous atoms. Moreover, each sample at different resolutions is assigned a different coding coefficient in the multi-dictionary learning process, so that the learned dictionary is more in line with the actual situation. Furthermore, a regularization term based on the fractional norm is drawn into the dictionary coupled learning to remove the redundant information in the dictionary, which can reduce the negative impacts of the redundant information. Comprehensive results demonstrate that the proposed JCRHR method achieves better results than the state-of-the-art methods, on several small sample face databases.

Application of the random dichotomous classification method to assessment builder’s brand based on consumer’s reviews

The brand influences the success and sustainability of the builder. In this paper, the authors continue to address the problem of managing developer branding based on information on the Internet. The paper proposes a dichotomous classification that allows you to classify developer reviews based on randomly labeled data. he dichotomous classification method allows you to work with small databases, in contrast to methods based on the construction of artificial neural networks, which require large samples of data. The method was implemented in the "Eidos" system. The Eidos system is based on automated system cognitive analysis (ASC-analysis). The decision to change the recall marker was made on the basis of changing the credibility of the model by Van Riesbergen's F-measure. The application of the dichotomous classification method to classify developers' reviews was tested on the example of OJSC "Perm Silicate Panels Plant" (PZSP). The dichotomous classification method allowed us to distinguish 20 negative and 63 positive reviews in the collected sample. The performance and adequacy of the method used is shown. In particular, it is shown that it is possible to divide the collected textual information about the developer into two clusters without prior training. The dichotomous classification method reduces the developer's response time to a reference. This paper proposes an algorithm for reference clustering based on the dichotomous classification method. This algorithm can be used as a basis for software to collect and analyze developer data, as well as to manage developer branding.

Increased maternal non-oxidative energy metabolism mediates association between prenatal di-(2-ethylhexyl) phthalate (DEHP) exposure and offspring autism spectrum disorder symptoms in early life: A birth cohort study

Prenatal phthalate exposure has previously been linked to the development of autism spectrum disorder (ASD). However, the underlying biological mechanisms remain unclear. We investigated whether maternal and child central carbon metabolism is involved as part of the Barwon Infant Study (BIS), a population-based birth cohort of 1,074 Australian children. We estimated phthalate daily intakes using third-trimester urinary phthalate metabolite concentrations and other relevant indices. The metabolome of maternal serum in the third trimester, cord serum at birth and child plasma at 1 year were measured by nuclear magnetic resonance. We used the Small Molecule Pathway Database and principal component analysis to construct composite metabolite scores reflecting metabolic pathways. ASD symptoms at 2 and 4 years were measured in 596 and 674 children by subscales of the Child Behavior Checklist and the Strengths and Difficulties Questionnaire, respectively. Multivariable linear regression analyses demonstrated (i) prospective associations between higher prenatal di-(2-ethylhexyl) phthalate (DEHP) levels and upregulation of maternal non-oxidative energy metabolism pathways, and (ii) prospective associations between upregulation of these pathways and increased offspring ASD symptoms at 2 and 4 years of age. Counterfactual mediation analyses indicated that part of the mechanism by which higher prenatal DEHP exposure influences the development of ASD symptoms in early childhood is through a maternal metabolic shift in pregnancy towards non-oxidative energy pathways, which are inefficient compared to oxidative metabolism. These results highlight the importance of the prenatal period and suggest that further investigation of maternal energy metabolism as a molecular mediator of the adverse impact of prenatal environmental exposures such as phthalates is warranted.