Training Dataset Research Articles

Predicting ignitability classification of thermally thick solids using hybrid GA-BPNN and PSO-BPNN algorithms

Two hybrid binary classification models combining the merits of GA (genetic algorithm) and PSO (particle swarm optimization) with BPNN (backpropagation neural network) are proposed to estimate the ignitability of thermally thick solids. A 1D numerical model verified by analytical correlation and experimental measurements is employed to yield training, validation and testing datasets. Surface absorptivity (ε), density (ρ), specific heat (Cp), thermal conductivity (k), critical temperature (Tcri), and incident heat flux (HF) which are highly related to solid ignition are selected as inputs, and ignitability serves as output. A one-hidden-layer BPNN is first constructed, and then GA and PSO are encoded to form GA-BPNN and PSO-BPNN where connection weights and biases are optimized by GA/PSO and the predicted error serves as objective function of GA/PSO. Results show that the optimal neuron number in hidden layer is 34. Compared to BPNN, GA-BPNN and PSO-BPNN feature higher accuracy but more training time. Accuracy and convergence efficiency of PSO-BPNN are higher than GA-BPNN, nevertheless GA-BPNN exhibits better robustness. Predictive accuracies of BPNN, GA-BPNN, and PSO-BPNN on testing dataset are 99.05%, 99.55%, and 99.55%, respectively. ROC (receiver operating characteristic) curves are plotted to rank the performance of the three models, revealing 0.5 is an appropriate threshold to classify ignitability. Feature importance of the models is analyzed using PI (permutation importance) and MIV (mean impact value) methods. It is found HF and Tcri are the most important two inputs followed by ε, while ρ, Cp, and k show least and approximately identical impact. Reliability of the three models is verified by predicting critical heat flux of ignition of three polymers, and good agreement with literature is observed.

Prediction of human O-linked glycosylation sites using stacked generalization and embeddings from pre-trained protein language model.

O-linked glycosylation, an essential post-translational modification process in Homo sapiens, involves attaching sugar moieties to the oxygen atoms of serine and/or threonine residues. It influences various biological and cellular functions. While threonine or serine residues within protein sequences are potential sites for O-linked glycosylation, not all serine and/or threonine residues undergo this modification, underscoring the importance of characterizing its occurrence. This study presents a novel approach for predicting intracellular and extracellular O-linked glycosylation events on proteins, which are crucial for comprehending cellular processes. Two base multi-layer perceptron models were trained by leveraging a stacked generalization framework. These base models respectively employ ProtT5 and Ankh O-linked glycosylation site-specific embeddings whose combined predictions are used to train the meta-multi-layer perceptron model. Trained on extensive O-linked glycosylation datasets, the stacked-generalization model demonstrated high predictive performance on independent test datasets. Furthermore, the study emphasizes the distinction between nucleocytoplasmic and extracellular O-linked glycosylation, offering insights into their functional implications that were overlooked in previous studies. By integrating the protein language model's embedding with stacked generalization techniques, this approach enhances predictive accuracy of O-linked glycosylation events and illuminates the intricate roles of O-linked glycosylation in proteomics, potentially accelerating the discovery of novel glycosylation sites. Stack-OglyPred-PLM produces Sensitivity, Specificity, Matthews Correlation Coefficient, and Accuracy of 90.50%, 89.60%, 0.464, and 89.70%, respectively on a benchmark NetOGlyc-4.0 independent test dataset. These results demonstrate that Stack-OglyPred-PLM is a robust computational tool to predict O-linked glycosylation sites in proteins. The developed tool, programs, training, and test dataset are available at https://github.com/PakhrinLab/Stack-OglyPred-PLM. Supplementary data are available at Bioinformatics online.

Automated tree crown labeling with 3D radiative transfer modelling achieves human comparable performances for tree segmentation in semi-arid landscapes

Mapping tree crowns in arid or semi-arid areas, which cover around one-third of the Earth’s land surface, is a key methodology towards sustainable management of trees. Recent advances in deep learning have shown promising results for tree crown segmentation. However, a large amount of manually labeled data is still required. We here propose a novel method to delineate tree crowns from high resolution satellite imagery using deep learning trained with automatically generated labels from 3D radiative transfer modeling, intending to reduce human annotation significantly. The methodological steps consist of 1) simulating images with a 3D radiative transfer model, 2) image style transfer learning based on generative adversarial network (GAN) and 3) tree crown segmentation using U-net segmentation model. The delineation performances of the proposed method have been evaluated on a manually annotated dataset consisting of more than 40,000 tree crowns. Our approach, which relies solely on synthetic images, demonstrates high segmentation accuracy, with an F1 score exceeding 0.77 and an Intersection over Union (IoU) above 0.64. Particularly, it achieves impressive accuracy in extracting crown areas (r2 greater than 0.87) and crown densities (r2 greater than 0.72), comparable to that of a trained dataset with human annotations only. In this study, we demonstrated that the integration of a 3D radiative transfer model and GANs for automatically generating training labels can achieve performances comparable to human labeling, and can significantly reduce the time needed for manual labeling in remote sensing segmentation applications.

73 Predictors of neurodevelopmental impairment in extreme preterm infants with unremarkable cranial ultrasound: A Canadian population-based cohort

Abstract Background Extreme preterm infants are at high-risk of brain injury and neurodevelopmental impairment (NDI). Previous studies reported significant NDI (sNDI) in preterm infants with no abnormalities on routine cranial ultrasound screening (CUS), traditionally perceived to have favorable outcomes. Recent data from population-based outlining clinical predictors of sNDI in those infants are lacking. Objectives (1) to describe the incidence, perinatal and neonatal characteristics of NDI and sNDI in a large population-based cohort of extreme preterm infants with unremarkable CUS; (2) to develop a prediction model of sNDI (any of: CP ≥ GMFCS stage 3, Bayley III < 70 in any domain, deafness requiring aids, or bilateral blindness) in those infants. Design/Methods A retrospective review of a national cohort of preterm infants born at <29 weeks’ gestation between April 2009 and December 2018. We included surviving infants with unremarkable routine CUS (no hemorrhage, periventricular leukomalacia or ventricular dilatation) who had neurodevelopmental assessment at 18-24 month corrected age at one of participating centres in the Canadian Neonatal Follow Up Network. We compared the baseline characteristics between infants with no NDI and those with any NDI and sNDI. The population sample was then randomly split into training and testing datasets in 70:30 ratio for prediction model’s development and validation, respectively. Multivariate logistic regression with GEE accounting for clustering within site was used, adjusted for variables with p-value<0.1 using backward selection procedure. The AUC and the diagnostic properties of the model developed using training dataset was then validated in the testing dataset. Results Out of 7463 eligible infants, 4327 (58%) were included. Of those, 2501 (59%) had no NDI and 1736 (41%) had NDI. sNDI accounted for 14% of the cohort and one third of those with NDI. Perinatal and neonatal characteristics of the three groups are shown in Table 1. On internal validation, maternal hypertension, maternal level of education, gestational age, male sex, ROP≥ stage 3 and systemic steroids were predictive of sNDI (Table 2). The prediction model had Sensitivity 0.91, NPV 0.90, and AUC 0.69 (95% CI 0.64-0.73). Conclusion In this national cohort, 2 out of 5 extreme preterm infants with unremarkable CUS had NDI and 1 in 7 had sNDI. The developed prediction model had high sensitivity and NPV, but poor discrimination. The findings of this study can help clinicians to identify at-risk infants who may benefit from early targeted interventions.

Ensemble LVQ Model for Photovoltaic Line-to-Line Fault Diagnosis Using K-Means Clustering and AdaGrad

Line-to-line (LL) faults are one of the most frequent short-circuit conditions in photovoltaic (PV) arrays which are conventionally detected and cleared by overcurrent protection devices (OCPDs). However, OCPDs are shown to face challenges when detecting LL faults under critical detection conditions, i.e., low mismatch levels and/or high fault impedance values. This occurs due to insufficient fault current, thus leaving the LL faults undetected and leading to power losses and even catastrophic fire hazards. To compensate for OCPD deficiencies, recent studies have proposed modern artificial intelligence (AI)-based methods. However, various limitations can still be witnessed even in AI-based methods, such as (i) most of the models requiring a massive training dataset, (ii) critical fault detection conditions not being taken into consideration, (iii) models not being accurate enough when dealing with critical conditions, etc. To this end, the present paper proposes a learning vector quantization (LVQ)-based ensemble learning model in which three LVQs are individually trained to detect and classify LL faults in PV arrays. The initial LVQ vectors are determined using the k-means clustering method, and the learning rate is optimized by the adaptive gradient (AdaGrad) optimizer. The training and testing datasets are collected according to the PV array’s current–voltage (I–V) characteristic curve, and several features are extracted based on the Canberra and chi-squared distance techniques. The model utilizes a small training dataset, considers various critical detection conditions for LL faults—such as different mismatch levels and fault impedance values—and the final experimental results show that the model achieves an impressive average accuracy of 99.26%.

Predicting financial performance with intellectual capital using machine learning

Purpose This study aims to apply machine learning techniques to efficiently predict leisure firms’ financial performance. Accurate financial forecasting is crucial in leisure and tourism, greatly affecting firms’ strategic decisions and competitive positioning. This study emphasizes the roles of intellectual capital to offer a nuanced understanding of how these types of capital influence firm success. Design/methodology/approach Using comprehensive firm-level data, this study examines several machine learning algorithms’ predictive capacity across a spectrum of industry sectors (general, manufacturing, service) to identify the most effective model and training dataset. These tools are used to evaluate financial metrics such as return on sales, return on assets and sales growth. A range of variables are incorporated into this process to enhance model accuracy and relevance. Findings Results demonstrate the support vector machine algorithm’s exceptional performance based on a training data set from the service sector in predicting leisure firms’ return on sales and sales growth. This algorithm is thus an efficacious strategic forecasting instrument. The variables significantly affecting firm performance include demand variation; organizational, product and technological innovation; synergistic innovation between multiple domains; salary levels; market strategy; and the number of employees. Originality/value By integrating advanced machine learning techniques with the strategic management of intellectual capital, this study presents a sophisticated approach to predicting leisure firms’ financial performance. Findings enrich the discourse on firm performance forecasting and offer actionable insights into strategic planning and resource allocation for practitioners in the leisure and tourism sectors.

Admission Prediction for Freshmen Students Using Low-cost Data Analytics Tool

This research paper tackles the admission prediction of enrollment for the tertiary level using the Naïve Bayes classification technique. Using data mining, the Naïve Bayes algorithm is simulated using Data Analytics Tool (DAT) and Weka data mining to provide the classifier model in evaluating if the given college freshmen applicant will continue or discontinue to be admitted to college. In line with the method, the researcher used gender, senior high school average grade, senior high school type, senior high school strand, and college entrance exam result. The training data set used for this machine learning model comprises 750 records from the different freshmen applications. As a result of the classifier model testing in Weka, the correct instances captured from the original 750 records is 86.27%, which is significantly higher and makes the classifier model reliable. In addition, a confusion matrix is performed and computed with an accuracy of 86.27%, precision of 83.13%, recall of 95.34%, and F1 score of 88.82%. Thus, the model test result is favorable to the reliability of the classifier model using Naïve Bayes technique.

Prediction of a Cephalometric Parameter and Skeletal Patterns from Lateral Profile Photographs: A Retrospective Comparative Analysis of Regression Convolutional Neural Networks.

Background/Objectives: Cephalometric analysis has a pivotal role in the quantification of the craniofacial skeletal complex, facilitating the diagnosis and management of dental malocclusions and underlying skeletal discrepancies. This study aimed to develop a deep learning system that predicts a cephalometric skeletal parameter directly from lateral profile photographs, potentially serving as a preliminary resource to motivate patients towards orthodontic treatment. Methods: ANB angle values and corresponding lateral profile photographs were obtained from the medical records of 1600 subjects (1039 female and 561 male, age range 3 years 8 months to 69 years 1 month). The lateral profile photographs were randomly divided into a training dataset (1250 images) and a test dataset (350 images). Seven regression convolutional neural network (CNN) models were trained on the lateral profile photographs and measured ANB angles. The performance of the models was assessed using the coefficient of determination (R2) and mean absolute error (MAE). Results: The R2 values of the seven CNN models ranged from 0.69 to 0.73, and the MAE values ranged from 1.46 to 1.53. Among the seven models, InceptionResNetV2 showed the highest success rate for predictions of ANB angle within 1° of range and the highest performance in skeletal class prediction, with macro-averaged accuracy, precision, recall, and F1 scores of 73.1%, 78.5%, 71.1%, and 73.0%, respectively. Conclusions: The proposed deep CNN models demonstrated the ability to predict a cephalometric skeletal parameter directly from lateral profile photographs, with 71% of predictions being within 2° of accuracy. This level of accuracy suggests potential clinical utility, particularly as a non-invasive preliminary screening tool. The system's ability to provide reasonably accurate predictions without radiation exposure could be especially beneficial for initial patient assessments and may enhance efficiency in orthodontic workflows.

Rapid deposition analysis of inhaled aerosols in human airways

A rapid data-driven method for determining regional deposition of inhaled medication aerosols in human airways is presented, which is patient specific. Inhalation patterns, device characteristics, and aerodynamic particle size distribution of medications are considered. The method is developed using dimensional analysis and Buckingham Pi theorem, and provides total, regional, and lobar distributions of aerosol deposition. 34 dimensionless quantities are selected, of which 22 encode features of the airway trees and segmented lobes, 14 pertain to the device and the drug formulation, and 13 the inhalation profile of the subject. The dimensionless correlations are obtained using a large database of computational fluid dynamics results on patient specific airways. The intraclass correlation coefficient between the current method and its training dataset is 0.92. The difference between the predicted average lobar deposition in the six asthma patients and the in-vivo data is 1.3%. The model has the potential to offer insights into the effectiveness of personalized drug delivery in clinical settings and can aid in drug development cycles.

Predicting interfacial tension in brine-hydrogen/cushion gas systems under subsurface conditions: Implications for hydrogen geo-storage

Underground hydrogen storage (UHS) critically relies on cushion gas to maintain pressure balance during injection and withdrawal cycles, prevent excessive water inflow, and expand storage capacity. Interfacial tension (IFT) between brine and hydrogen/cushion gas mixtures is a key factor affecting fluid dynamics in porous media. This study develops four machine learning models— Decision Trees (DT), Random Forests (RF), Support Vector Machines (SVM), and Multi-Layer Perceptrons (MLP)—to predict IFT under geo-storage conditions. These models incorporate variables such as pressure, temperature, molality, overall gas density, and gas composition to evaluate the impact of different cushion gases. A group-based data splitting method enhances the realism of our tests by preventing information leakage between training and testing datasets. Shapley Additive Explanations (SHAP) reveal that while the MLP model prioritizes gas composition, the RF model focuses more on operational parameters like pressure and temperature, showing distinct predictive dynamics. The MLP model excels, achieving coefficients of determination (R2) of 0.96, root mean square error (RMSE) of 2.10 mN/m, and average absolute relative deviation (AARD) of 3.25%. This robustness positions the MLP model as a reliable tool for predicting IFT values between brine and hydrogen/cushion gas (es) mixtures beyond the confines of the studied dataset. The findings of this study present a promising approach to optimizing hydrogen geo-storage through accurate predictions of IFTs, offering significant implications for the advancement of energy storage technologies.

79 Using artificial intelligence and machine learning to automate injury prevention surveillance data

Abstract Background The Canadian Hospitals Injury Reporting and Prevention Program (CHIRPP) is a pan-Canadian injury and poisoning surveillance system that tracks patients presenting to Canadian Emergency Departments (EDs). Like many surveillance systems, data is input in a manual fashion with humans screening all charts. Natural language processing algorithms using transformer models can analyze text from triage notes of previous CHIRPP-positive patients to train models to predict future CHIRPP patients. We sought to automate these functions using machine learning and artificial intelligence. Objectives The objective of this initiative is to automate detection of injured patients using artificial intelligence. Once charts are identified, a further summarization model will automatically generate a narrative of the events surrounding the injury, as required for the surveillance system. Design/Methods ED notes were obtained from an electronic health record for all patients visiting the ED from October 2018 to December 2021 (n =203626). All charts were manually reviewed to determine if patients met the criteria for CHIRPP. CHIRPP-positive patients made up 24.8% (n = 50499) of the total presentations. Patient charts were randomly split into train (70%) and validation (30%) datasets. Pre-trained distilbert-base-uncased model from Huggingface was utilized. The best model according to AUROC was selected for further development. For note summarization, we fine-tuned a multi-task model (t5-small) from the Huggingface repository. We used human summaries from the same dataset (October 2018-December 2021) as training data (49,078 note, summary pairs). A 70/30 train/validation split was performed. Results Our model captured 90% of all positive cases while reducing the total number of charts for review by 77%, 75% and 63% for train, validation and test sets respectively. See AUC curve (figure 1). Based on 2022 volumes this is a reduction to 24,027 from 64,938 charts. For note summarization, when we compared the cosine similarities of human summaries to actual note text vs t5-small summaries to note text we saw remarkable similarity implying high model performance. Conclusion Our model has shown that high volumes of charts that were previously screened manually can utilize natural language processing to automate the task with minimal sacrifices in terms of data capture and accuracy, dramatically reducing cost while improving speed to reporting. Our summarization model also showed potential for further automating this process. This application to an injury prevention program has wide scale applicability and scalability.

Machine Learning Modeling for Predicting Tensile Strain Capacity of Pipelines and Identifying Key Factors

Abstract Machine learning (ML) techniques have recently gained great attention across a multitude of engineering domains, including pipeline materials. However, their application to tensile strain capacity (TSC) modeling remains unexplored. To bridge this gap, this study developed and evaluated an ML model to predict the tensile strain capacity of girth-welded pipelines. The model was trained on over 20,000 data points derived from a TSC equation available in the literature. The ML model demonstrated robust performance in predicting tensile strain capacities. Evidence of this lies in the near-zero means, minimal standard deviations, and normal distribution of residuals for both the training and test datasets. These collectively suggest that the model provides a good fit for the data. Furthermore, the model's loss behavior indicates successful convergence and generalization, without signs of overfitting or underfitting. An analysis using the random forest method revealed that the geometry of the flaw, specifically the flaw depth, is the most influential variable in predicting the TSC. This could be attributed to its significant impact on the fracture toughness of materials. In contrast, material properties and fracture toughness exert less influence relatively, despite their contributions to the model. This finding underscores the importance of flaw geometry in TSC prediction models. Overall, the development of a data-driven TSC model has shown efficient TSC modeling. This model leverages ML techniques, allowing for continuous updates with new data via deep learning.

A Novel Interpolation Method for Soil Parameters Combining RBF Neural Network and IDW in the Pearl River Delta

Soil fertility is a critical factor in agricultural production, directly impacting crop growth, yield, and quality. To achieve precise agricultural management, accurate spatial interpolation of soil parameters is essential. This study developed a new interpolation prediction framework that combines Radial Basis Function (RBF) neural networks with Inverse Distance Weighting (IDW), termed the IDW-RBFNN. This framework initially uses the IDW method to apply preliminary weights based on distance to the data points, which are then used as input for the RBF neural network to form a training dataset. Subsequently, the RBF neural network further trains on these data to refine the interpolation results, achieving more precise spatial data interpolation. We compared the interpolation prediction accuracy of the IDW-RBFNN framework with ordinary Kriging (OK) and RBF methods under three different parameter settings. Ultimately, the IDW-RBFNN demonstrated lower error rates in terms of RMSE and MRE compared to direct RBF interpolation methods when adjusting settings based on different power values, even with a fixed number of data samples. As the sample size decreases, the interpolation accuracy of OK and RBF methods is significantly affected, while the error of IDW-RBFNN remains relatively low. Considering both interpolation accuracy and resource limitations, we recommend using the IDW-RBFNN method (p = 2) with at least 60 samples as the minimum sampling density to ensure high interpolation accuracy under resource constraints. Our method overcomes limitations of existing approaches that use fixed steady-state distance decay parameters, providing an effective tool for soil fertility monitoring in delta regions.

W1-Net: a highly scalable ptychography convolutional neural network

X-ray ptychography is a coherent diffraction imaging technique that allows for the quantitative retrieval of both the amplitude and phase information of a sample in diffraction-limited resolution. However, traditional reconstruction algorithms require a large number of iterations to obtain phase and amplitude images exactly, and the expensive computation precludes real-time imaging. To solve the inverse problem of ptychography data, PtychoNN uses deep convolutional neural networks for real-time imaging. However, its model is relatively simple, and its accuracy is limited by the size of the training dataset, resulting in lower robustness. To address this problem, a series of W-Net neural network models have been proposed which can robustly reconstruct the object phase information from the raw data. Numerical experiments demonstrate that our neural network exhibits better robustness, superior reconstruction capabilities and shorter training time with high-precision ptychography imaging.

Predictive Regression Modeling for Forecasting Graduation Duration in Online Offsite Degree Program

The demand for Information Technology (IT) professionals continues to rise across various sectors, where theyplay vital roles. However, the supply of IT graduates often fails to meet industry needs and this is a huge problem for the SriLankan IT Industry (National IT-BPM Workforce Survey – 2019). In this context, this study presents a predictive regressionmodelling approach to predict graduation duration in the Bachelor of Information Technology (BIT) degree program at theUniversity of Moratuwa, Sri Lanka. It integrates demographic data—student district, birth year, AL results, OL maths grade,gender, employability status, occupation, and AL stream—along with academic performance indicators like diplomacompletions and higher diploma completions. After evaluating the suggested features, the key findings indicate thesignificance of certain features, notably the number of semesters taken to complete the diploma, higher diploma, and thedegree. Additionally, demographic factors such as district, birth year, AL results, OL maths grade, gender, and employabilitystatus were found to be important. The regression analysis was carried out using the Orange data mining tool (Orange DataMining). Various algorithms, including random forest, neural network, linear regression, and k-nearest neighbours (kNN),were used to develop predictive models. By adjusting parameters such as metrics, weights, number of neighbours, numberof iterations, and training dataset size, the models were optimised to better fit the dataset. Training and testing the modelsrevealed consistent error metrics, including MSE, RMSE, MAE, and R^2, validating the accuracy of predictions. Byconsidering the least and reasonable error in each model, the most suitable model to fit the given dataset was selected.The prediction model accurately forecasted graduation duration for subsequent academic batches, demonstrating itseffectiveness in predicting student progress in the program. This research contributes to understanding the factorsinfluencing graduation duration in a distance learning context and provides insights for educational institutions to optimiseprogram planning and student support initiatives. Additionally, it is a good indicator to the companies to gain a betterunderstanding of the availability of future workforce.

Detection of hypertension from pharyngeal images using deep learning algorithm in primary care settings in Japan

BackgroundThe early detection of hypertension using simple visual images in a way that does not require physical interaction or additional devices may improve quality of care in the era of...

Backdoor Attacks in Peer-to-Peer Federated Learning

Most machine learning applications rely on centralized learning processes, opening up the risk of exposure of their training datasets. While federated learning (FL) mitigates to some extent these privacy risks, it relies on a trusted aggregation server for training a shared global model. Recently, new distributed learning architectures based on Peer-to-Peer Federated Learning (P2PFL) offer advantages in terms of both privacy and reliability. Still, their resilience to poisoning attacks during training has not been investigated. In this paper, we propose new backdoor attacks for P2PFL that leverage structural graph properties to select the malicious nodes, and achieve high attack success, while remaining stealthy. We evaluate our attacks under various realistic conditions, including multiple graph topologies, limited adversarial visibility of the network, and clients with non-IID data. Finally, we show the limitations of existing defenses adapted from FL and design a new defense that successfully mitigates the backdoor attacks, without an impact on model accuracy.

GCS-YOLOv8: A Lightweight Face Extractor to Assist Deepfake Detection.

To address the issues of target feature blurring and increased false detections caused by high compression rates in deepfake videos, as well as the high computational resource requirements of existing face extractors, we propose a lightweight face extractor to assist deepfake detection, GCS-YOLOv8. Firstly, we employ the HGStem module for initial downsampling to address the issue of false detections of small non-face objects in deepfake videos, thereby improving detection accuracy. Secondly, we introduce the C2f-GDConv module to mitigate the low-FLOPs pitfall while reducing the model's parameters, thereby lightening the network. Additionally, we add a new P6 large target detection layer to expand the receptive field and capture multi-scale features, solving the problem of detecting large-scale faces in low-compression deepfake videos. We also design a cross-scale feature fusion module called CCFG (CNN-based Cross-Scale Feature Fusion with GDConv), which integrates features from different scales to enhance the model's adaptability to scale variations while reducing network parameters, addressing the high computational resource requirements of traditional face extractors. Furthermore, we improve the detection head by utilizing group normalization and shared convolution, simplifying the process of face detection while maintaining detection performance. The training dataset was also refined by removing low-accuracy and low-resolution labels, which reduced the false detection rate. Experimental results demonstrate that, compared to YOLOv8, this face extractor achieves the AP of 0.942, 0.927, and 0.812 on the WiderFace dataset's Easy, Medium, and Hard subsets, representing improvements of 1.1%, 1.3%, and 3.7% respectively. The model's parameters and FLOPs are only 1.68 MB and 3.5 G, reflecting reductions of 44.2% and 56.8%, making it more effective and lightweight in extracting faces from deepfake videos.

Development of an evaluation method for precipitation particle types by using disdrometer data

Abstract A method for evaluating precipitation particle types from the size-fall velocity data observed by disdrometers by combining an expectation-maximization (EM) algorithm and a self-organizing map (SOM) was developed. An EM algorithm was used to estimate the particle size-fall velocity relationships according to previous work, and the SOM map was used to classify the relationships into graupel or snow categories. The method was applied for snowfall data observed by the volume scanning video disdrometer at a site in Sapporo, Japan, and the daily graupel-to-solid precipitation ratio (G/S ratio) from 1 December, 2017 to 28 February, 2018 was obtained. The sensitivity of the results to different SOM configurations, SOM node groupings, and training datasets was also tested. The G/S ratio obtained from the simulation data of a meteorological model was compared as an example of the application of the product created by the new method.

Precision phenotyping from routine laboratory parameters for out of hospital survival prediction in an all comers prospective PCI registry

Out-of-hospital mortality in coronary artery disease (CAD) is particularly high and established adverse event prediction tools are yet to be available. Our study aimed to investigate whether precision phenotyping can be performed using routine laboratory parameters for the prediction of out-of-hospital survival in a CAD population treated by percutaneous coronary intervention (PCI). All patients treated by PCI and discharged alive in a tertiary center between January 2016 – December 2022 that have been included prospectively in the local registry were analyzed. 115 parameters from the PCI registry and 266 parameters derived from routine laboratory testing were used. An extreme gradient-boosted decision tree machine learning (ML) algorithm was trained and used to predict all-cause and cardiovascular-cause survival. A total of 4027 patients with 4981 PCI hospitalizations were randomly included in the 70% training dataset and 1729 patients with 2160 PCI hospitalizations were randomly included in the 30% validation dataset. All-cause and cardiovascular cause mortality was 17.5% and 12.2%. The integrated area under the receiver operator characteristic curve for prediction of all-cause and cardiovascular cause mortality by the ML on the validation dataset was 0.844 and 0.837, respectively (all p < 0.001). Parameters reflecting renal function (first and maximum serum creatinine), hematologic function (mean corpuscular hemoglobin concentration, platelet distribution width), and inflammatory status (lymphocyte per monocyte ratio) were among the most important predictors. Accurate out-of-hospital survival prediction in CAD can be achieved using routine laboratory parameters. ML outperformed clinical risk scores in predicting out-of-hospital mortality in a prospective all-comers PCI population and has the potential to precisely phenotype patients.