Sparse Data Research Articles

Correction to: A roadmap for generating annual bycatch estimates from sparse at-sea observer data

Correction to: A roadmap for generating annual bycatch estimates from sparse at-sea observer data

Are the postmortem concentration changes of the synthetic cannabinoid cumyl-5F-P7AICA and its N-pentanoic acid metabolite dependent on the environmental conditions? – A systematic study following pulmonary administration to pigs

The number of fatal cases involving synthetic cannabinoids (SCs) is increasing. However, interpretation of postmortem (PM) toxicological findings is challenging, due to unknown PM intervals and possible redistribution processes or instabilities. Only sparse data on SCs are available. Therefore, a controlled pig study was performed to determine the PM stability of cumyl-5F-P7AICA under different environmental conditions. Ten pigs inhalatively received 200 µg/kg body weight cumyl-5F-P7AICA each. Six hours later, they were euthanized and biopsies of the main tissues and body fluids were taken. Animals were stored in air or water (n=5 each) and samples were repeatedly taken for three days. Quantification of cumyl-5F-P7AICA and its N-pentanoic acid metabolite (NPA) was performed using standard addition or a fully validated method (blood) followed by LC-MS/MS. Time-dependent concentration changes of cumyl-5F-P7AICA were observed in liver, bile fluid and muscle specimens at both conditions. Concentrations of NPA only changed considerably in duodenum content of animals stored in air. Environment-dependent concentrations changes were only observed for cumyl-5F-P7AICA in kidney and the NPA metabolite in duodenum content. Overall, cumyl-5F-P7AICA and its metabolite seem to be quite stable in PM specimens. Hence, also central blood might be analyzed, if no peripheral blood is available.

SS4CTR: a semi-supervised framework for enhancing click-through rate prediction in sparse and imbalanced data

SS4CTR: a semi-supervised framework for enhancing click-through rate prediction in sparse and imbalanced data

Wind turbine dynamic wake flow estimation (DWFE) from sparse data via reduced-order modeling-based machine learning approach

Wind turbine wake poses a significant challenge in wind farm operations, affecting power generation efficiency. This study introduces a Dynamic Wake Flow Estimation (DWFE) framework designed to predict wind turbine wake evolution from sparse measurement data. The framework integrates Gaussian Process Regression (GPR), Proper Orthogonal Decomposition (POD), and Long Short-Term Memory (LSTM) networks. Specifically, GPR plays a pivotal role in DWFE by enabling the transformation of flow fields discrete sensor signals into coherent, low-dimensional flow fields which is crucial for accurate flow field predictions, while POD aids in dimensionality reduction and LSTM forecasts temporal wake dynamics, improving both predictive accuracy and computational efficiency. Parametric analysis demonstrates that the robustness and adaptability of the framework improve prediction accuracy with increased sensor density and flexible POD modes. The DWFE framework demonstrates high accuracy in wake flow estimation even with limited data, effectively capturing dynamic wake behavior. Future work will extend this approach to various topographic and climatic conditions, optimizing computational efficiency, and improving interpretability. These advancements will expand the framework's applicability in wind energy and other engineering fields requiring precise flow prediction, emphasizing DWFE's potential in advancing wind farm design, operation and renewable energy optimization.

Research on a three-dimensional radiation field reconstruction algorithm based on an improved 3D CNN

This article investigates the application of an improved three-dimensional convolutional neural network (3D CNN) for sparse data-based reconstruction of radiation fields. Sparse radiation data points are consolidated into structured three-dimensional matrices and fed into a self-attention integrated CNN, enabling the network to interpolate and produce complete radiation distribution grids. The model’s validity is assessed through experiments with randomly sourced radiation in scenarios both with and without shielding, as well as in refined grid configurations. Results indicate that in unshielded environments, a mere 5%(15 points) sampling yields an average relative error of 4%, while in shielded settings, a 7% (21 points) sampling maintains the error around 11%. In refined grid contexts, a 2% sampling rate suffices to limit the error to 6.58%. Thus, the improved 3D CNN is demonstrated to be highly effective for precise three-dimensional radiation field reconstruction in sparse data scenarios.

Tree-based classification model for Long-COVID infection prediction with age stratification using data from the National COVID Cohort Collaborative.

We propose and validate a domain knowledge-driven classification model for diagnosing post-acute sequelae of SARS-CoV-2 infection (PASC), also known as Long COVID, using Electronic Health Records (EHRs) data. We developed a robust model that incorporates features strongly indicative of PASC or associated with the severity of COVID-19 symptoms as identified in our literature review. The XGBoost tree-based architecture was chosen for its ability to handle class-imbalanced data and its potential for high interpretability. Using the training data provided by the Long COVID Computation Challenge (L3C), which was a sample of the National COVID Cohort Collaborative (N3C), our models were fine-tuned and calibrated to optimize Area Under the Receiver Operating characteristic curve (AUROC) and the F1 score, following best practices for the class-imbalanced N3C data. Our age-stratified classification model demonstrated strong performance with an average 5-fold cross-validated AUROC of 0.844 and F1 score of 0.539 across the young adult, mid-aged, and older-aged populations in the training data. In an independent testing dataset, which was made available after the challenge was over, we achieved an overall AUROC score of 0.814 and F1 score of 0.545. The results demonstrated the utility of knowledge-driven feature engineering in a sparse EHR data and demographic stratification in model development to diagnose a complex and heterogeneously presenting condition like PASC. The model's architecture, mirroring natural clinician decision-making processes, contributed to its robustness and interpretability, which are crucial for clinical translatability. Further, the model's generalizability was evaluated over a new cross-sectional data as provided in the later stages of the L3C challenge. The study proposed and validated the effectiveness of age-stratified, tree-based classification models to diagnose PASC. Our approach highlights the potential of machine learning in addressing the diagnostic challenges posed by the heterogeneity of Long-COVID symptoms.

Spatiotemporal upscaling of sparse air-sea pCO2 data via physics-informed transfer learning.

Global measurements of ocean pCO2 are critical to monitor and understand changes in the global carbon cycle. However, pCO2 observations remain sparse as they are mostly collected on opportunistic ship tracks. Several approaches, especially based on direct learning, have been used to upscale and extrapolate sparse point data to dense estimates using globally available input features. However, these estimates tend to exhibit spatially heterogeneous performance. As a result, we propose a physics-informed transfer learning workflow to generate dense pCO2 estimates that are grounded in real-world measurements and remainphysically consistent. The models are initially trained on dense input predictors against pCO2 estimates from Earth system model simulation, and then fine-tuned to sparse SOCAT observational data. Compared to the benchmark direct learning approach, our transfer learning framework shows major improvements of up to 56-92%. Furthermore, we demonstrate that using models that explicitly account for spatiotemporal structures in the data yield better validation performances by 50-68%. Our strategy thus presents a new monthly global pCO2 estimate that spans for 35 years between 1982-2017.

Predicting anthropogenic food supplementation from individual tracking data

Many wildlife species consume food or refuse provided by humans. To understand the effect of anthropogenic food subsidies on wildlife populations, we first need to quantify where and when individuals can access such food sources. The Red Kite Milvus milvus is an opportunistic raptor species and uses both inadvertent and deliberate food subsidies provided by citizens. Here we present a new approach using global positioning system (GPS)‐tracking data to predict where anthropogenic food subsidies probably occur. We tracked 497 individuals with solar‐powered GPS transmitters over an average of 3.2 (range 1–9) breeding seasons in Switzerland, and combined these data with locations of 125 known feeding sites obtained through interviews. We used two sequential random forest models, at both individual movement and population levels, to predict where anthropogenic food subsidies were attended by Red Kites. The first model classified locations that were frequently and regularly revisited, and successfully predicted 85% of locations that were within 50 m of an externally validated feeding site. These predicted locations were aggregated in 500‐m grid cells to calculate the proportion of individuals and locations associated with predicted food subsidy. A second model related the presence of known food subsidies to the aggregated predictions. In our study area, 80% of known anthropogenic food provision locations could be correctly identified using Red Kite tracking data, but data sparsity beyond the core range of tracked individuals limits predictions of anthropogenic food subsidies at larger geographical scales. Nonetheless, biologging data can identify ephemeral food sources, and facilitate an assessment of the importance of anthropogenic food subsidies for the fitness of individuals in tracked populations.

Enabling personalized smart tourism with location-based social networks

With the rapid advance of mobile internet, communication technology and the Internet of Things (IoT), the tourism industry is undergoing unprecedented transformation. Smart tourism offers users personalized and customized services for travel planning and recommendations. Location-based social networks (LBSNs) play a crucial role in smart tourism industry by providing abundant data sources through their social networking attributes. However, applying LBSNs to smart tourism is a challenge due to the need to deal with complex multi-source information modeling and tourism data sparsity. In this article, to fully harness the potential of LBSNs using deep learning technologies, we propose an knowledge-driven personalized recommendation method for smart tourism. Representation learning techniques can effectively modeling the contextual information (e.g., time, space, and semantics) in LBSNs, while the data augmentation strategy of contrastive learning techniques can explore user personalized travel behaviors and alleviate data sparsity. To demonstrate the effectiveness of the proposed approach, we conducted a case study on trip recommendation. Furthermore, the patterns of human mobility are revealed by exploring the effect of contextual data and tourist potential preferences.

Local patterns of spread of influenza A H3N2 virus in coastal Kenya over a 1-year period revealed through virus sequence data.

The patterns of spread of influenza A viruses in local populations in tropical and sub-tropical regions are unclear due to sparsity of representative spatiotemporal sequence data. We sequenced and analyzed 58 influenza A(H3N2) virus genomes sampled between December 2015 and December 2016 from nine health facilities within the Kilifi Health and Demographic Surveillance System (KHDSS), a predominantly rural region, covering approximately 891km2 along the Kenyan coastline. The genomes were compared with 1571 contemporaneous global sequences from 75 countries. We observed at least five independent introductions of A(H3N2) viruses into the region during the one-year period, with the importations originating from Africa, Europe, and North America. We also inferred 23 virus location transition events between the nine facilities included in the study. International virus imports into the study area were captured at the facilities of Chasimba, Matsangoni, Mtondia, and Mavueni, while all four exports from the region were captured from the Chasimba facility, all occurring to Africa destinations. A strong spatial clustering of virus strains at all locations was observed associated with local evolution. Our study shows that influenza A(H3N2) virus epidemics in local populations appear to be characterized by limited introductions followed by significant local spread and evolution. Knowledge of the viral lineages that circulate within specific populations in understudied tropical and subtropical regions is required to understand the full diversity and global ecology of influenza viruses and to inform vaccination strategies within these populations.

Use of three-dimensional crossgradient joint inversion of marine controlled-source electromagnetic (CSEM) and seismic refraction data for overburden geohazards mapping.

Offshore energy-transition activities typically require the construction of new subsurface facilities where overburden heterogeneity and deformation pattern may not be fully understood. Anisotropic marine controlled-source electromagnetic (CSEM) and seismic refraction imaging can be adapted to geohazards assessment of the overburden to mitigate safety and costs risks in such areas. This is investigated using an area in deepwater NW Borneo as the type-example. For this proof-of-concept study, a well-established Multiphysics joint inversion code was extended to handle marine seismic refraction data from conventional towed streamers. Realistic field-scale synthetic models with anomalous resistivities and velocities in the shallow section (∼50-200 m below mudline) were built and their CSEM and seismic refraction responses were first inverted separately and then jointly in 3D using the crossgradient method. The joint inversion recovered the anomalous bodies better than the separate inversions. The velocity model from the joint inversion showed improvement from the separate inversion but required a good starting model. As a further test of the methodology, first-break travel times were picked from legacy 3D seismic data and inverted with the available sparse CSEM data. For model verification, resistivity and velocity profiles from separate and joint inversion models were extracted at a well location in the survey area and compared with actual well logs. The joint inversion models matched the well logs better than the separate inversion models. Moreover, known geohazards (complex fault-sets and lateral-inflows) from independent geomatic study were imaged in the joint inversion models. Also, the resistivity anisotropy ratio tracked anomalous zones that correspond to zones of recent complex faulting and potential lateral inflows in the overburden mass transport systems. Thus, joint inversion of CSEM and seismic refraction data has potential to improve the mapping of overburden hazards offshore, with the associated resistivity anisotropy serving as a proxy for active ground deformation.

Seal Pipeline: Enhancing Dynamic Object Detection and Tracking for Autonomous Unmanned Surface Vehicles in Maritime Environments

This study addresses the dynamic object detection problem for Unmanned Surface Vehicles (USVs) in marine environments, which is complicated by boat tilting and camera illumination sensitivity. A novel pipeline named “Seal” is proposed to enhance detection accuracy and reliability. The approach begins with an innovative preprocessing stage that integrates data from the Inertial Measurement Unit (IMU) with LiDAR sensors to correct tilt-induced distortions in LiDAR point cloud data and reduce ripple effects around objects. The adjusted data are grouped using clustering algorithms and bounding boxes for precise object localization. Additionally, a specialized Kalman filter tailored for maritime environments mitigates object discontinuities between successive frames and addresses data sparsity caused by boat tilting. The methodology was evaluated using the VRX simulator, with experiments conducted on the Volga River using real USVs. The preprocessing effectiveness was assessed using the Root Mean Square Error (RMSE) and tracking accuracy was evaluated through detection rate metrics. The results demonstrate a 25% to 30% improvement in detection accuracy and show that the pipeline can aid industry even with sparse object representation across different frames. This study highlights the potential of integrating sensor fusion with specialized tracking for accurate dynamic object detection in maritime settings, establishing a new benchmark for USV navigation systems’ accuracy and reliability.

A GAN-based stepwise full-field mechanical prediction model for architected metamaterials

The exploration of mechanical metamaterials, characterized by unique unit cells with significant macroscopic mechanical properties, is of crucial importance. Traditional methods relying on bioinspired structures lack broad design space and controllability to some extent, especially when practical requirements have no biological counterparts. Hence, there is a pressing need to advance methodologies for the expansive design of unit cells, catering to diverse practical demands. Inspired by architected material, a latent design paradigm gains prominence for its extensive possibilities and unique distribution. However, wide design space faces challenges in accurately assessing the deformation history and stress states of diverse units, requiring validation through experiments or simulations, and imposing efficiency constraints on design. Using patterns generated by a mimetic corrosion algorithm and computational simulations, a GAN-based machine learning surrogate model is constructed to predict the history of deformation and stress fields of architected unit cells with accuracy and efficiency, in which a stepwise mapping strategy is proposed to augment comprehension of mechanical information in image processing. In distinct cases covering both explored domains and several unexplored domains, the model achieves effective predictions, demonstrating the robustness and adaptability of the model in extrapolating predictions of both time domain and design space. The proposed method demonstrates efficacy, stability, and generalizability, pioneering a solution for the mechanical history prediction of architected metamaterials in a sparse data setting, thereby overcoming the limitations of the training set.

Compilation of Shape Operators on Sparse Arrays

We show how to build a compiler for a sparse array language that supports shape operators such as reshaping or concatenating arrays, in addition to compute operators. Existing sparse array programming systems implement generic shape operators for only some sparse data structures, reduce shape operators on other data structures to those, and do not support fusion. Our system compiles sparse array expressions to code that efficiently iterates over reshaped views of irregular sparse data structures, without needing to materialize temporary storage for intermediates. Our evaluation shows that our approach generates sparse array code competitive with popular sparse array libraries: our generated shape operators achieve geometric mean speed-ups of 1.66×–15.3× when compared to hand-written kernels in scipy.sparse and 1.67×–651× when compared to generic implementations in pydata/sparse. For operators that require data structure conversions in these libraries, our generated code achieves geometric mean speed-ups of 7.29×–13.0× when compared to scipy.sparse and 21.3×–511× when compared to pydata/sparse. Finally, our evaluation demonstrates that fusing shape and compute operators improves the performance of several expressions by geometric mean speed-ups of 1.22×–2.23×.

The Analytics of Robust Satisficing: Predict, Optimize, Satisfice, Then Fortify

In the paper, “The Analytics of Robust Satisficing: Predict, Optimize, Satisfice, Then Fortify,” published in Operations Research, authors Sim, Tang, Zhou, and Zhu introduce a novel approach to decision making under uncertainty. Their method, termed “estimation-fortified robust satisficing,” leverages advanced predictive and prescriptive analytics to optimize decisions where traditional models falter due to risk ambiguity and estimation uncertainties. This approach not only enhances the resilience of decisions against unforeseen variations but also consistently outperforms conventional predictive methods in scenarios characterized by sparse data. This significant advancement promises to fortify decision-making processes in critical sectors such as finance and operations management, offering a new paradigm in handling the inherent uncertainties of real-world systems.

Non-linear dynamics of global liquidity and energy sector profitability

ABSTRACT This study examines how global liquidity influences the profitability of companies in the energy sector using a non-linear approach. Applying Functional Principal Component Analysis (FPCA) to sparse data, we reconstruct the profitability history of 497 energy companies in various subsectors, including coal, oil and gas, oil and gas-related equipment and services, renewable energy, and uranium. We use a Distributed Delay Nonlinear Model (DLNM) to estimate the impact of global liquidity on profitability. Our results reveal distinct nonlinear patterns in the response of these subsectors to changes in global liquidity. For example, in the oil and gas subsector, an extreme quantile (99%) of global liquidity is associated with a significant 5.19% increase in profitability in the first quarter after the shock. In contrast, in the renewable energy subsector, a lower quantile (25%), corresponding to a moderately downward trend, is associated with a 0.73% increase in profitability over the same period. These finings are crucial to improving risk management, investment and policy development strategies within the energy sector, offering a deeper understanding of the dynamics in the global financial and economic landscape.

Development of brain tumor radiogenomic classification using GAN-based augmentation of MRI slices in the newly released gazi brains dataset

Significant progress has been made recently with the contribution of technological advances in studies on brain cancer. Regarding this, identifying and correctly classifying tumors is a crucial task in the field of medical imaging. The disease-related tumor classification problem, on which deep learning technologies have also become a focus, is very important in the diagnosis and treatment of the disease. The use of deep learning models has shown promising results in recent years. However, the sparsity of ground truth data in medical imaging or inconsistent data sources poses a significant challenge for training these models. The utilization of StyleGANv2-ADA is proposed in this paper for augmenting brain MRI slices to enhance the performance of deep learning models. Specifically, augmentation is applied solely to the training data to prevent any potential leakage. The StyleGanv2-ADA model is trained with the Gazi Brains 2020, BRaTS 2021, and Br35h datasets using the researchers’ default settings. The effectiveness of the proposed method is demonstrated on datasets for brain tumor classification, resulting in a notable improvement in the overall accuracy of the model for brain tumor classification on all the Gazi Brains 2020, BraTS 2021, and Br35h datasets. Importantly, the utilization of StyleGANv2-ADA on the Gazi Brains 2020 Dataset represents a novel experiment in the literature. The results show that the augmentation with StyleGAN can help overcome the challenges of working with medical data and the sparsity of ground truth data. Data augmentation employing the StyleGANv2-ADA GAN model yielded the highest overall accuracy for brain tumor classification on the BraTS 2021 and Gazi Brains 2020 datasets, together with the BR35H dataset, achieving 75.18%, 99.36%, and 98.99% on the EfficientNetV2S models, respectively. This study emphasizes the potency of GANs for augmenting medical imaging datasets, particularly in brain tumor classification, showcasing a notable increase in overall accuracy through the integration of synthetic GAN data on the used datasets.

Enhanced knowledge graph recommendation algorithm based on multi-level contrastive learning

Integrating the Knowledge Graphs (KGs) into recommendation systems enhances personalization and accuracy. However, the long-tail distribution of knowledge graphs often leads to data sparsity, which limits the effectiveness in practical applications. To address this challenge, this study proposes a knowledge-aware recommendation algorithm framework that incorporates multi-level contrastive learning. This framework enhances the Collaborative Knowledge Graph (CKG) through a random edge dropout method, which constructs feature representations at three levels: user-user interactions, item-item interactions and user-item interactions. A dynamic attention mechanism is employed in the Graph Attention Networks (GAT) for modeling the KG. Combined with the nonlinear transformation and Momentum Contrast (Moco) strategy for contrastive learning, it can effectively extract high-quality feature information. Additionally, multi-level contrastive learning, as an auxiliary self-supervised task, is jointly trained with the primary supervised task, which further enhances recommendation performance. Experimental results on the MovieLens and Amazon-books datasets demonstrate that this framework effectively improves the performance of knowledge graph-based recommendations, addresses the issue of data sparsity, and outperforms other baseline models across multiple evaluation metrics.

Structure of the basement of the near-Laptev part of the Eurasian basin according to geological and geophysical data

The near-Laptev part of the Eurasian basin is the area of transition from modern spreading to intraplate rifting. We propose an approach to the construction of a 3D model of the basement of the Laptev Sea part of the Eurasian basin based on the linkage of all currently available geological and geophysical data. The structure of the acoustic basement of the Eurasian basin is characterized by alternating troughs and highs in cross-section. The sparse seismic data do not allow us to directly trace the strike of these structures, but they can be correlated with the linearity established by gravity and magnetic data and related to the sequential opening of the basin. For the Laptev Sea part of the Eurasian basin, where linearity is no longer traceable from magnetic data, we propose a method of determining the strike of the basement structures on the basis of seismic stratigraphic analysis. The new 3D model of the acoustic basement in the studied area provided the basis for the tectonic scheme of the entire Eurasian basin. The model reflects the main stages of basin development: continental rifting up to 56 Ma, normal spreading 56–45 Ma, ultra-slow spreading 45–34 Ma, ultra-ultra-slow spreading 34–20 Ma. The southern part of the study area is overlain by a sedimentary cover with an age of 20 Ma and younger, which is associated with the cessation of spreading here no later than 20 Ma.

Prediction of Fan Array Performance with Polynomial and Support Vector Regression Models

The increasing utilisation of demand-controlled ventilation strategies leads to the frequent operation of fans under part-load conditions. To accurately predict the energy demand of a ventilation system with a fan array in the early design stages, models that calculate reliable results across the whole operating range are required. We present the comparison of a polynomial and a machine learning approach through support vector regression (SVR) to predict the fan performance over a wide range of typical operating points. For fitting and validation, we use experimental data. We investigate the extrapolation performance of both approaches. The SVR model achieves a slightly better representation of the experimental data with a lower error, especially when only sparse data are available. Both approaches yield similar results when the evaluation is conducted within the experimentally captured domain but deviates outside the domain. At operating points that are far from the experimentally captured domain, the polynomial models yield fan efficiencies that are physically plausible, while the SVR models drastically overpredict the fan efficiency. To rate the influence of such deviations towards modelling the actual energy demand, both approaches are applied to an operation simulation of a simplified office building. Both approaches yield similar results despite differing extrapolation capabilities.