Improve Model Accuracy Research Articles

Reinforcement learning-based assimilation of the WOFOST crop model

Crop model assimilation is a crucial technique for improving the accuracy and precision of crop models by integrating observational data and crop models. Although conventional data assimilation methods such as Kalman filtering and variational methods have been widely applied, these methods often face limitations in data quality, model bias, and high computational complexity. This study explored the potential of reinforcement learning (RL) for crop model assimilation, which has the advantage of not requiring large datasets. Based on the WOFOST crop model, two RL environments were constructed: a Daily-Data Driven approach and a Time-Series Driven approach. The Proximal Policy Optimization (PPO) algorithm was used to train these environments for 100,000 iterations. The assimilation results were compared with the commonly used SUBPLEX optimization algorithm using four-year field measurement data and a public dataset with added random errors. Our results demonstrate that the Time-Series Driven RL model achieved assimilation accuracy comparable to the SUBPLEX optimization algorithm, with an average MAE of 0.65 compared to 0.76 for SUBPLEX, and a slight decrease in RMSE, while significantly reducing the computational burden by 365 times. In a multi-year stability test, the Time-Series Driven RL model and SUBPLEX had similar assimilation performance. This study demonstrates the potential of RL for crop model assimilation, providing a novel approach to overcome the limitations of conventional assimilation algorithms. The findings suggest that RL-based crop model assimilation can improve model accuracy and efficiency, with potential for practical applications in precision agriculture.

Fusion of Raman and FTIR Spectroscopy Data Uncovers Physiological Changes Associated with Lung Cancer.

Due to the high mortality rate, more effective non-invasive diagnostic methods are still needed for lung cancer, the most common cause of cancer-related death worldwide. In this study, the integration of Raman and Fourier-transform infrared spectroscopy with advanced data-fusion techniques is investigated to improve the detection of lung cancer from human blood plasma samples. A high statistical significance was found for important protein-related oscillations, which are crucial for differentiating between lung cancer patients and healthy controls. The use of low-level data fusion and feature selection significantly improved model accuracy and emphasizes the importance of structural protein changes in cancer detection. Although other biomolecules such as carbohydrates and nucleic acids also contributed, proteins proved to be the decisive markers found using this technique. This research highlights the power of these combined spectroscopic methods to develop a non-invasive diagnostic tool for discriminating lung cancer from healthy state, with the potential to extend such studies to a variety of other diseases.

A Novel YOLOv5_ES based on lightweight small object detection head for PCB surface defect detection

In the manufacturing process of printed circuit boards (PCBs), surface defects have a significant negative impact on product quality. Considering that traditional object detection algorithms have low accuracy in handling PCB images with complex backgrounds, various types, and small-sized defects, this paper proposes a PCB defect detection algorithm based on a novel YOLOv5 multi-scale attention mechanism(EMA) spatial pyramid dilated Convolution (SPD-Conv) (YOLOv5_ES) network improved YOLOv5s framework. Firstly, the detection head is optimized by removing medium and large detection layers, fully leveraging the capability of the small detection head to identify minor target defects. This approach not only improves model accuracy but also achieves lightweighting. Secondly, in order to further reduce the number of parameters and computational costs, the SPD-Conv is introduced to improve the feature extraction capability by reducing information loss. Thirdly, a EMA module is introduced to fuse context information of different scales, enhancing the model’s generalization ability. Compared to the YOLOv5s model, there is a 3.1% improvement in mean average precision (mAP0.5), a 55.8% reduction in model parameters, and a 4.8% reduction in giga floating-point operations per second (GFLOPs). These results demonstrate a significant improvement in both accuracy and model parameter efficiency.

Integrating solar-induced chlorophyll fluorescence with traditional remote sensing and environmental variables for enhanced rice yield prediction in Nepal using machine learning

Rice is a vital cereal crop providing essential food for Nepal. To mitigate climate risks and ensure food security, especially with climate change and increasing extreme weather events, it is imperative and a prerequisite to have timely and reliable rice yield prediction. While traditional vegetation indices (VIs) and environmental variables are commonly used in rice yield prediction, the potential of solar-induced chlorophyll fluorescence (SIF), despite its greater sensitivity, has been explored in limited studies. Furthermore, integrating SIF with other remote sensing and environmental data for regional crop yield modeling still needs to be explored. This study aims to address these gaps by incorporating SIF to enhance rice yield prediction in Nepal, offering a novel approach to improving model accuracy. This study uses multi-source data and machine learning algorithms to study 22 major rice-growing districts in the Terai region of Nepal. Random forest (RF), support vector machine (SVR), gradient boosting regressor (GBR), and least absolute shrinkage and selection operator (LASSO) algorithms were employed to predict crop yield using various combinations of input features. Machine learning models were trained and tested over four different time windows within the growing period of rice, using 15 years of rice yield data from 2003 to 2017. Our analysis reveals that GBR out-performed other machine learning algorithms when using input features from July to September, achieving root mean square error (RMSE) and the ratio of RMSE (RRMSE) to observed mean values of 250 kg/ha and 8.6%, respectively. The findings also reveal that incorporating SIF alongside traditional remote sensing and environmental variables significantly enhances the model's prediction accuracy. These findings are valuable for ensuring food security, optimizing agricultural resource management, and supporting decision-making for farmers, policymakers, and supply chain stakeholders.

The Identification and Quantification of Hidden Hazards in Small Scale Reservoir Engineering Based on Deep Learning: Intelligent Perception for Safety of Small Reservoir Projects in Jiangxi Province

This study aims to enhance the detection and assessment of safety hazards in small-scale reservoir engineering using advanced image processing and deep learning techniques. Given the critical importance of small reservoirs in flood management, water supply, and ecological balance, the effective monitoring of their structural integrity is crucial. This paper developed a fully convolutional semantic segmentation method for hidden danger images of small reservoirs using an encoding–decoding structure, utilizing a deep learning framework of convolutional neural networks (CNNs) to process and analyze high-resolution images captured by unmanned aerial vehicles (UAVs). The method incorporated data augmentation and adaptive learning techniques to improve model accuracy under diverse environmental conditions. Finally, the quantification data of hidden dangers (length, width, area, etc.) were obtained by converting the image pixels to the actual size. Results demonstrate significant improvements in detecting structural deficiencies, such as cracks and seepage areas, with increased precision and recall rates compared to conventional methods, and the HHSN-25 network (Hidden Hazard Segmentation Network with 25 layers) proposed in this paper outperforms other methods. The main evaluation indicator, mIoU of HHSN-25, is higher than other methods, reaching 87.00%, and the Unet is 85.50%, and the Unet++ is 85.55%. The proposed model achieves reliable real-time performance, allowing for early warning and effective management of potential risks. This study contributes to the development of more efficient monitoring systems for small-scale reservoirs, enhancing their safety and operational sustainability.

Revealing the intricacies of natural convection: A key factor in aqueous zinc battery design

Aqueous zinc-based batteries offer high safety, abundant resources, low cost, and environmental friendliness, making them a promising alternative to lithium-ion batteries for large-scale energy storage markets. However, the natural convection in the electrolyte during operation, which can greatly affect the battery performance, is overlooked for a long time. Through particle tracing experiments, this study demonstrates that in the vertical placement of the battery, natural convection occurs due to the variation of electrolyte density, whereas it does not happen when the cathode is on top. Besides, there is a significant difference in electrochemical performance between these two configurations, with a voltage difference of up to 0.25 V at 10 mA cm⁻² and a peak current difference of up to 3.4 mA in linear sweep voltammetry. Simulation results show that convective mass transfer can be up to 100 times greater than diffusive mass transfer in a bulk solution. This highlights the importance of considering natural convection to improve model accuracy. Further, to increase energy density, batteries are scaled up in practical applications to reduce the weight of the casing. This work simulates mass transfer conditions under various scenarios, illustrating the trade-off between current density and battery size, and offering new guidance for the design of aqueous zinc-based batteries.

Bayesian inference of coupled groundwater flow and radiogenic helium-4 production and transport at the catchment scale

Hydrogeological numerical models are essential for assessing radioactive waste disposal by understanding groundwater flow systems. These models typically rely on hydraulic head data, with other state variables often underutilized in model inversions. In Flanders' Neogene aquifer, where safety studies for Boom Clay are ongoing, existing models face uncertainties due to dependence on hydraulic heads alone. This study aims to: i) develop a conceptual model of radiogenic Helium-4 (4Herad) production and transport to reproduce observed concentrations; ii) evaluate the usefulness of 4Herad observations as an additional state variable for inverse conditioning; and iii) assess how including 4Herad affects model calibration and uncertainty reduction. We address these objectives by incorporating radiogenic Helium-4 (4Herad) as an unconventional state variable in model inversion to tackle parameter non-uniqueness. Utilizing a Bayesian approach, we employ two 3D models: a groundwater flow (GWF) model conditioned by hydraulic heads and a coupled GWF model with 4Herad production and transport, conditioned by both hydraulic heads and 4Herad concentrations. We hypothesize that integrating 4Herad will improve model accuracy and reduce uncertainties compared to using hydraulic head data alone. Findings indicate that major 4Herad sources include crustal fluxes (~57 %) and in situ Boom Clay production (~25 %). The coupled inference narrows the posterior distributions of parameters, especially for hydraulic conductivity (e.g., Mol, Diest, Berchem formations, and the Rauw Fault) and recharge, showing that 4Herad observations effectively reduce uncertainty beyond hydraulic head data. Including 4Herad improves model reliability and prediction accuracy, highlighting its importance for refining groundwater flow parameters and derived fluxes. To the best of our knowledge, this study marks the first-ever attempt at harnessing 4Herad quantitative insights into flow and transport parameters following a Bayesian approach at a catchment scale, setting a precedent for future research and emphasizing the innovative nature of this work. This study showcases a state-of-the-art approach in hydrogeological modelling, demonstrating the effectiveness of integrating 4Herad with hydraulic head observations to produce more reliable model predictions. By reducing uncertainties in crucial groundwater fluxes, this approach offers significant potential for improving geological disposal safety studies. This pioneering work advocates for the continued use of 4Herad in hydrogeological modelling, emphasizing its innovative contribution to future research and safety studies. Plain language summaryUnderstanding how groundwater flows is crucial for safely storing radioactive waste. Traditionally, models used to study groundwater rely mostly on measurements of water pressure, while other valuable data often go unused. In the Neogene aquifer in Flanders, where researchers are studying the Boom Clay formation for potential waste storage, relying only on water pressure has created uncertainties in the models. This study explores a new approach by using a different type of measurement: radiogenic Helium-4 (4Herad). The goals of the study are threefold: first, to build a model that accurately represents how 4Herad is produced and moves through the groundwater; second, to test whether including 4Herad as an additional measurement improves the model compared to using only water pressure data; and third, to see how using 4Herad affects the accuracy and reliability of the model. The study uses two types of models: one that relies only on water pressure data and another that includes both water pressure and 4Herad measurements. The idea is that 4Herad could provide additional insights and help reduce uncertainties in the model. The findings show that 4Herad comes mainly from two sources: the surrounding crust and the Boom Clay itself. Including 4Herad in the model helps narrow down uncertainties, particularly regarding how water moves through different layers and how much water is being replenished. This research is the first to use 4Herad in this way, using a formal statistical method to improve groundwater models. The results suggest that 4Herad is a valuable addition for making more accurate predictions and improving safety assessments for radioactive waste storage.

FedDA: Resource-adaptive federated learning with dual-alignment aggregation optimization for heterogeneous edge devices

Federated learning (FL) is an emerging distributed learning paradigm that allows multiple clients to collaborate on training a global model without sharing their local data. However, in practical heterogeneous edge device scenarios, FL faces the challenges of system heterogeneity and data heterogeneity, which leads to unfair participation and degraded global model performance. In this paper, we introduce FedDA, a resource-adaptive FL framework, which adapts to the client’s computing resources by assigning heterogeneous models of different sizes. To improve the performance of heterogeneous model aggregation and adjust to non-independent and identically distributed (non-i.i.d.) data, we propose a dual-alignment aggregation optimization method, i.e., parameter feature space alignment and output space alignment. Specifically, FedDA exploits the permutation symmetry property of weight space to permutate the model parameter positions via an adaptive layer-wise matching method, thereby aligning models with significant deviations in parameter feature space. FedDA mitigates the imbalance in parameter quantity between smaller and larger models through parameter expansion. Additionally, FedDA maps client labels into a uniform embedding space through output space alignment, thus reducing model parameter deviations due to non-i.i.d. data without additional client-side computing overhead. We evaluate the performance of FedDA on benchmark datasets, including FashionMNIST, CIFAR10, CIFAR100 and AGNews. Experimental results demonstrate that FedDA achieves up to 8.71% improvement in model accuracy compared to baseline methods, highlighting its effectiveness in addressing the challenges of heterogeneity.

The most efficient machine learning algorithms in stroke prediction: A systematic review.

Stroke is one of the most common causes of death worldwide, leading to numerous complications and significantly diminishing the quality of life for those affected. The purpose of this study is to systematically review published papers on stroke prediction using machine learning algorithms and introduce the most efficient machine learning algorithms and compare their performance. The papers have published in period from 2019 to August 2023. The authors conducted a systematic search in PubMed, Scopus, Web of Science, and IEEE using the keywords "Artificial Intelligence," "Predictive Modeling," "Machine Learning," "Stroke," and "Cerebrovascular Accident" from 2019 to August 2023. Twenty articles were included based on the inclusion criteria. The Random Forest (RF) algorithm was introduced as the best and most efficient stroke ML algorithm in 25% of the articles (n = 5). In addition, in other articles, Support Vector Machines (SVM), Stacking and XGBOOST, DSGD, COX& GBT, ANN, NB, and RXLM algorithms were introduced as the best and most efficient ML algorithms in stroke prediction. This research has shown a rapid increase in using ML algorithms to predict stroke, with significant improvements in model accuracy in recent years. However, no model has reached 100% accuracy or is entirely error-free. Variations in algorithm efficiency and accuracy stem from differences in sample sizes, datasets, and data types. Further studies should focus on consistent datasets, sample sizes, and data types for more reliable outcomes.

FLAV: Federated Learning for Autonomous Vehicle privacy protection

Autonomous Vehicle Systems are committed to safer, more efficient and more convenient transportation on the roads of the future. However, concerns about vehicle data privacy and security remain significant. Federated Learning, as a decentralized machine learning approach, allows multiple devices or data sources to collaboratively train models without sharing raw data, providing essential privacy protection. In this paper, we propose a privacy-preserving framework for autonomous vehicles, named FLAV. First, we use a multi-chain parallel aggregation strategy to transmit model parameters and design a model parameter filtering mechanism, which reduces communication overhead by filtering out the local model parameters of certain vehicles, thereby alleviating bandwidth pressure. Second, we introduce a dynamic adjustment mechanism that automatically adjusts the regularization strength in the loss function by comparing the differences between each vehicle’s local model parameters and the cumulative parameters from the preceding vehicles in the chain. This mechanism effectively balances local model training with global model consistency, ensuring the model’s adaptability to local data while enhancing the coordination of models between vehicles in the chain. Experimental results demonstrate that our proposed method reduces communication costs while improving model accuracy and privacy protection, effectively ensuring the security of autonomous driving data.

High-fidelity model development of CO2 booster refrigeration systems in supermarkets using field measurements

Supermarket refrigeration systems adopting traditional refrigerants with high global warming potential (GWP) have impacts on global warming for indirect and direct greenhouse gases (GHG) emissions. CO2 is a popular low-GWP alternative. The transcritical operation of CO2 systems worsens their energy performance, but provides recoverable heat as a heat source to reduce gas consumption. To evaluate operation performance, data-driven models, trained by historical data, are weak in implementation with datasets outside the scope of training data; in contrast, theoretical models have better extrapolation ability to calculate all operation conditions of CO2 systems at supermarket. Existing theoretical modeling approaches often lack validation against the limited public-access data, which reduces model reliability for further applications, and adopt oversimplified inference methods for unmeasured variables, which increases the risks of breaking thermodynamic laws and lowering model accuracy. This study therefore develops a steady-state theoretical model for CO2 booster refrigeration systems validated against field measurements from three UK supermarkets. The available measurements are utilized to the best level to ensure model accuracy and physical interpretability. Proposed methods to infer missing variables in CO2 systems include condenser outlet temperature, evaporating temperature, compressor isentropic efficiency and compressor mass flow rate. Results show that proposed inference methods enhance the abilities of the proposed modelling approach to ensure data integrity, avoid breaking thermodynamic laws, and improve model accuracy by reflecting real-time actual values of unmeasured variables rather than rough assumptions. The proposed modeling approach provides satisfactory accuracy validated using high-resolution measurements across the whole year from three real supermarkets.

FedADMM-InSa: An inexact and self-adaptive ADMM for federated learning

Federated learning (FL) is a promising framework for learning from distributed data while maintaining privacy. The development of efficient FL algorithms encounters various challenges, including heterogeneous data and systems, limited communication capacities, and constrained local computational resources. Recently developed FedADMM methods show great resilience to both data and system heterogeneity. However, they still suffer from performance deterioration if the hyperparameters are not carefully tuned. To address this issue, we propose an inexact and self-adaptive FedADMM algorithm, termed FedADMM-InSa. First, we design an inexactness criterion for the clients’ local updates to eliminate the need for empirically setting the local training accuracy. This inexactness criterion can be assessed by each client independently based on its unique condition, thereby reducing the local computational cost and mitigating the undesirable straggle effect. The convergence of the resulting inexact ADMM is proved under the assumption of strongly convex loss functions. Additionally, we present a self-adaptive scheme that dynamically adjusts each client’s penalty parameter, enhancing algorithm robustness by mitigating the need for empirical penalty parameter choices for each client. Extensive numerical experiments on both synthetic and real-world datasets have been conducted. As validated by some tests, our FedADMM-InSa algorithm improves model accuracy by 7.8% while reducing clients’ local workloads by 55.7% compared to benchmark algorithms.

Evaluating crop nitrogen status in maize leaves: A predictive modelling approach using chlorophyll fluorescence parameters

The consumption of chemical fertilizers has increased eight-fold since the 19th century, outstripping crop yields increases and, emphasizing the need for precise nitrogen (N) assessment in crops to optimize fertilization and mitigate environmental impacts. This study developed a model using chlorophyll fluorescence technology to accurately evaluate the N status in maize leaves while addressing the limitations of current labor-intensive and environmentally sensitive methods. Based on a long-term experiment initiated in 2011, maize hybrid Fumin 985 was sampled in 2021 and 2022 under two crop-straw management strategies (SM: no tillage with surface straw mulch, SP: plough tillage with straw incorporation) and six N application rates. Partial least squares regression (PLSR) models were formulated using chlorophyll fluorescence parameters (ChlF) to assess leaf N content (N leaf). The results indicated that a N application rate of 270 kg ha−1 sufficed to meet crop N requirements. Leaf characteristics such as N leaf, total pigment content (TP), and leaf dry weight (DW leaf) changed significantly with increasing N application rates, influencing rapid chlorophyll fluorescence (OJIP) dynamics. Principal component analysis (PCA) reduced ChlF from 35 to 21, and four models were developed, among which, the model using ChlF and TP was more accurate than the model using DW alone. Key ChlF parameters for PLSR model performance included ABS/RC, φ(Eo), ETo/CSm, and δ(Ro)/(1–δ(Ro)). Although non-destructive N leaf detection using chlorophyll fluorescence technology proved feasible, additional leaf characteristics, such as TP, are necessary to improve model accuracy. Considering local field conditions is essential for the application of this technology at a larger scale. Precise evaluation of N status using chlorophyll fluorescence is beneficial for a more efficient N management and sustainable agriculture.

End-to-end AI pipeline optimization: Benchmarking and performance enhancement techniques for recommendation systems

End-to-end AI pipeline optimization is critical for improving the efficiency and performance of recommendation systems, which play a pivotal role in personalizing user experiences across various domains. This review explores benchmarking and performance enhancement techniques tailored to recommendation systems within AI pipelines. The objective is to streamline the processes involved in data ingestion, feature engineering, model training, and deployment to achieve optimal system performance and user satisfaction. Recommendation systems typically involve complex workflows that require continuous optimization. Benchmarking serves as a foundational step, enabling the identification of bottlenecks and inefficiencies within the pipeline. By establishing clear performance metrics, such as precision, recall, and latency, benchmarking allows for the comparative analysis of different algorithms, data processing methods, and system configurations. These metrics guide the selection of the most suitable models and techniques, thereby enhancing overall system effectiveness. Performance enhancement techniques are then applied to various stages of the AI pipeline. Advanced methods in feature engineering, such as automated feature selection and dimensionality reduction, can significantly improve model accuracy while reducing computational overhead. In the model training phase, techniques like hyperparameter tuning, gradient-based optimization, and distributed training are employed to accelerate convergence and improve model generalization. Additionally, optimization strategies at the deployment stage, including model compression, quantization, and the use of specialized hardware, are crucial for minimizing latency and resource consumption. This review also highlights the importance of continuous monitoring and feedback loops to maintain the effectiveness of recommendation systems in dynamic environments. By integrating real-time analytics and adaptive algorithms, systems can adjust to changing user behaviors and preferences, ensuring sustained performance improvements. In conclusion, optimizing end-to-end AI pipelines for recommendation systems involves a multifaceted approach that includes benchmarking, feature engineering, model training, and deployment enhancements. These efforts collectively contribute to more efficient, scalable, and accurate recommendation systems, ultimately leading to better user experiences and operational efficiencies. This paper will focus on optimizing AI pipelines for recommendation systems, covering the entire process from data extraction and feature engineering to model deployment and performance benchmarking. It will discuss techniques for identifying and mitigating performance bottlenecks in different computing environments, providing valuable insights for enhancing the efficiency of recommendation systems, which are crucial for various applications.

K-PCD: A new clustering algorithm for building energy consumption time series analysis and predicting model accuracy improvement

Clustering algorithms are often applied to building energy consumption data analysis to mine representative patterns of building energy usage. This paper proposes a new clustering algorithm: K-PCD, which is particularly suitable for building energy consumption time series. K-PCD utilizes a Pearson correlation coefficient-based distance measure (PCD), and a novel centroid calculation method that takes into account both the PCD and the traditional Euclidean distance (ED) between time series. This study also proposes a new clustering validity index (CVI) tailored to the predictability of building energy consumption: Energy Prediction Clustering Performance Index (EPCPI). The K-PCD and EPCPI are practically analyzed and validated using one year of data from 29 real buildings. The results indicate that comparing with traditional clustering algorithms, the K-PCD achieves better clustering results. This EPCPI index thoroughly explores building operation patterns, and after adding clustering labels, it can maximize the prediction accuracy of the Back Propagation Neural Network (BPNN) model for building energy consumption.

Comprehensive input models and machine learning methods to improve permeability prediction

This study investigates the use of machine learning techniques and the proper selection of input data to estimate permeability in geosciences, using six types of input logs: gamma ray (GR), resistivity (RT), effective porosity (PHIE), density (RHO), sonic (DT), and compensated neutron porosity (NPHI). A total of 57 models were constructed using combinations of these logs and tested using five machine learning methods: Extreme Learning Machine (ELM), Random Forest (RF), Gradient Boosting (GB), K-Nearest Neighbor (KNN), and Multilayer Perceptron (MLP). This approach produced 285 unique permeability predictions. RF had the highest correlation coefficient (0.925) and average error (0.196), indicating a precision-correlation trade-off. The ELM approach had the lowest average error, 0.083, and a correlation value of 0.871. Testing on a blind well revealed that the GB and RF approaches were highly effective in predicting permeability, with R² values of 0.92 and 0.90, respectively, even in untested settings. The findings emphasize the need of using appropriate machine learning algorithms and input data to improve model accuracy and reliability.

Local differential privacy federated learning based on heterogeneous data multi-privacy mechanism

Federated learning enables the development of robust models without accessing users data directly. However, recent studies indicate that federated learning remains vulnerable to privacy leakage. To address this issue, local differential privacy mechanisms have been incorporated into federated learning. Nevertheless, local differential privacy will reduce the availability of data. To explore the balance between privacy budgets and data availability in federated learning, we propose federated learning for clustering hierarchical aggregation with adaptive piecewise mechanisms under multiple privacy-FedAPCA as a way to balance the relationship between privacy preservation and model accuracy. First, we introduce an adaptive piecewise mechanism that dynamically adjusts perturbation intervals based on the data ranges across different layers of the model, ensuring minimized perturbation variance while maintaining the same level of privacy. Second, we propose two dynamic privacy budget allocation methods, which are allocating the privacy budget based on global accuracy and global loss, and allocating the privacy budget based on local accuracy and loss, to ensure that better model accuracy can be achieved under the same privacy budget. Finally, we propose a clustering hierarchical aggregation method in the model aggregation stage, and the model is updated and aggregated after the unbiased estimation of the disturbance in each cluster according to the variance of each layer. FedAPCA improves the balance between privacy preservation and model accuracy. Our experimental results, comparing FedAPCA with the SOTA multi-privacy local differential privacy federated learning frameworks on the MNIST and CIFAR-10 datasets, demonstrate that FedAPCA improves model accuracy by 1%–2%.

Machine learning radiomics based on intra and peri tumor PA/US images distinguish between luminal and non-luminal tumors in breast cancers

PurposeThis study aimed to evaluate a radiomics model using Photoacoustic/ultrasound (PA/US) imaging at intra and peri-tumoral area to differentiate Luminal and non-Luminal breast cancer (BC) and to determine the optimal peritumoral area for accurate classification. Materials and methodsFrom February 2022 to April 2024, this study continuously collected 322 patients at Shenzhen People’s Hospital, using standardized conditions for PA/US imaging of BC. Regions of interest were delineated using ITK-SNAP, with peritumoral regions of 2 mm, 4 mm, and 6 mm automatically expanded using code from the Pyradiomic package. Feature extraction was subsequently performed using Pyradiomics. The study employed Z-score normalization, Spearman correlation for feature correlation, and LASSO regression for feature selection, validated through 10-fold cross-validation. The radiomics model integrated intra and peri-tumoral area, evaluated by receiver operating characteristic curve(ROC), Calibration and Decision Curve Analysis(DCA). ResultsWe extracted and selected features from intratumoral and peritumoral PA/US images regions at 2 mm, 4 mm, and 6 mm. The comprehensive radiomics model, integrating these regions, demonstrated enhanced diagnostic performance, especially the 4 mm model which showed the highest area under the curve(AUC):0.898(0.78–1.00) and comparably high accuracy (0.900) and sensitivity (0.937). This model outperformed the standalone clinical model and combined clinical-radiomics model in distinguishing between Luminal and non-Luminal BC, as evidenced in the test set results. ConclusionThis study developed a radiomics model integrating intratumoral and peritumoral at 4 mm region PA/US model, enhancing the differentiation of Luminal from non-Luminal BC. It demonstrated the diagnostic utility of peritumoral characteristics, reducing the need for invasive biopsies and aiding chemotherapy planning, while emphasizing the importance of optimizing tumor surrounding size for improved model accuracy.

Incorporating Social Determinants of Health in Infectious Disease Models: A Systematic Review of Guidelines.

Infectious disease (ID) models have been the backbone of policy decisions during the COVID-19 pandemic. However, models often overlook variation in disease risk, health burden, and policy impact across social groups. Nonetheless, social determinants are becoming increasingly recognized as fundamental to the success of control strategies overall and to the mitigation of disparities. To underscore the importance of considering social heterogeneity in epidemiological modeling, we systematically reviewed ID modeling guidelines to identify reasons and recommendations for incorporating social determinants of health into models in relation to the conceptualization, implementation, and interpretations of models. After identifying 1,372 citations, we found 19 guidelines, of which 14 directly referenced at least 1 social determinant. Age (n = 11), sex and gender (n = 5), and socioeconomic status (n = 5) were the most commonly discussed social determinants. Specific recommendations were identified to consider social determinants to 1) improve the predictive accuracy of models, 2) understand heterogeneity of disease burden and policy impact, 3) contextualize decision making, 4) address inequalities, and 5) assess implementation challenges. This study can support modelers and policy makers in taking into account social heterogeneity, to consider the distributional impact of infectious disease outbreaks across social groups as well as to tailor approaches to improve equitable access to prevention, diagnostics, and therapeutics. Infectious disease (ID) models often overlook the role of social determinants of health (SDH) in understanding variation in disease risk, health burden, and policy impact across social groups.In this study, we systematically review ID guidelines and identify key areas to consider SDH in relation to the conceptualization, implementation, and interpretations of models.We identify specific recommendations to consider SDH to improve model accuracy, understand heterogeneity, estimate policy impact, address inequalities, and assess implementation challenges.

FDP-FL: differentially private federated learning with flexible privacy budget allocation

Abstract Federated learning (FL) as a privacy-preserving technology enables multiple clients to collaboratively train models on decentralized data. However, transmitting model parameters between local clients and the central server can potentially result in information leakage. Differentially private federated learning (DPFL) has emerged as a promising solution to enhance privacy. Nevertheless, existing DPFL schemes suffer from two issues: (i) most schemes that aim to achieve desired model accuracy may incur a high privacy budget. (ii) several schemes that consider the trade-off between privacy and accuracy by utilizing linear clipping bound may distort numerous model parameters. In this paper, we first propose FDP-FL, a flexible differential privacy approach for FL. FDP-FL introduces a novel series sum privacy budget allocation instead of uniform allocation and enables adaptive and nonlinear noise scale decay. In this way, a tight bound for cumulative privacy loss can be achieved while optimizing model accuracy. Then in order to mitigate gradient leakages caused by honest-but-curious clients and server, we further design client-level FDP-FL and record-level FDP-FL, respectively. Experimental results demonstrate that our FDP-FL improves model accuracy by $\sim $13.3% compared with the basic DP-FL under a fixed privacy budget and outperforms existing trade-off schemes with the same hyperparameter setting.