Data-centric Approach Research Articles

Seismic Soundoff: How AI is being applied to seismic interpretation

In episode 234 of SEG's Seismic Soundoff podcast, computer scientist Morten Ofstad discusses the use of artificial intelligence (AI) in seismic interpretation, focusing on the advantages of a data-centric approach over the traditional model-centric method.

Machine Learning for Failure Management in Microwave Networks: A Data-Centric Approach

Machine Learning for Failure Management in Microwave Networks: A Data-Centric Approach

A data-centric approach for ethical and trustworthy AI in journalism

AI-driven journalism refers to various methods and tools for gathering, verifying, producing, and distributing news information. Their potential is to extend human capabilities and create new forms of augmented journalism. Although scholars agreed on the necessity to embed journalistic values in these systems to make AI systems accountable, less attention was paid to data quality, while the results’ accuracy and efficiency depend on high-quality data in any machine learning task. Assessing data quality in the context of AI-driven journalism requires a broader and interdisciplinary approach, relying on the challenges of data quality in machine learning and the ethical challenges of using machine learning in journalism. To better identify these, we propose a data quality assessment framework to support the collection and pre-processing stages in machine learning. It relies on three of the core principles of ethical journalism—accuracy, fairness, and transparency—and participates in the shift from model-centric to data-centric AI, by focusing on data quality to reduce reliance on large datasets with errors, making data labelling consistent, and better integrating journalistic knowledge.

A Data-Centric Approach to improve performance of deep learning models

The Artificial Intelligence has evolved and is now associated with Deep Learning, driven by availability of vast amount of data and computing power. Traditionally, researchers have adopted a Model-Centric Approach, focusing on developing new algorithms and models to enhance performance without altering the underlying data. However, Andrew Ng, a prominent figure in the AI community, has recently emphasized on better (quality) data rather than better models, which has given birth to Data Centric Approach, also known as Data Oriented technique. The transition from model oriented to data oriented approach has rapidly gained momentum within the realm of deep learning. Despite its promise, the Data-Centric Approach faces several challenges, including (a) generating high-quality data, (b) ensuring data privacy, and (c) addressing biases to achieve fairness in datasets. Currently, there has been limited effort in preparing quality data. Our work aims to address this gap by focusing on the generation of high-quality data through methods such as data augmentation, multi-stage hashing to eliminate duplicate instances, to detect and correct noisy labels, using confident learning. The experiments on popular datasets, namely MNIST, Fashion MNIST, and CIFAR-10 were performed by utilizing ResNet-18 as the common framework followed by both Model Centric and Data Centric Approach. Comparative performance analysis revealed that the Data Centric Approach consistently outperformed the Model Centric Approach by a relative margin of at least 3%. This finding highlights the potential for further exploration and adoption of the Data-Centric Approach in various domains such as healthcare, finance, education, and entertainment, where the quality of data could significantly enhance the performance.

Critical insights into data curation and label noise for accurate prediction of aerobic biodegradability of organic chemicals.

The focus of this work is to enhance state-of-the-art Machine Learning (ML) models that can predict the aerobic biodegradability of organic chemicals through a data-centric approach. To do that, an already existing dataset that was previously used to train ML models was analyzed for mismatching chemical identifiers and data leakage between test and training set and the detected errors were corrected. Chemicals with high variance between study results were removed and an XGBoost was trained on the dataset. Despite extensive data curation, only marginal improvement was achieved in the classification model's performance. This was attributed to three potential reasons: (1) a significant number of data labels were noisy, (2) the features could not sufficiently represent the chemicals, and/or (3) the model struggled to learn and generalize effectively. All three potential reasons were examined and point (1) seemed to be the most decisive one that prevented the model from generating more accurate results. Removing data points with possibly noisy labels by performing label noise filtering using two other predictive models increased the classification model's balanced accuracy from 80.9% to 94.2%. The new classifier is therefore better than any previously developed classification model for ready biodegradation. The examination of the key characteristics (molecular weight of the substances, proportion of halogens present and distribution of degradation labels) and the applicability domain indicate that no/not a large share of difficult-to-learn substances has been removed in the label noise filtering, meaning that the final model is still very robust.

Machine Learning Approach to Investigate High Temperature Corrosion of Critical Infrastructure Materials

Degradation of coatings and structural materials due to high temperature corrosion in the presence of molten salt environment is a major concern for critical infrastructure applications to meet its commercial viability. The choice of high value coatings and structural (construction parts) materials comes with challenges, and therefore data centric approach may accelerate change in discovery and data practices. This research aims to use machine learning (ML) approach to estimate corrosion rates of materials when operated at high temperatures conditions (e.g., nuclear, geothermal, oxidation (dry/wet), solar applications) but geared towards nuclear thermochemical cycles. Published data related to materials (structural and coatings materials), their composition and manufacturing, including corrosion environment were gathered and analysed. Analysis demonstrated that random forest regression model is highly precise compared to other models. Assessment indicates that very limited sets of materials are likely to survive high temperature corrosive environment for extended period of exposure. While a higher quality and larger dataset are required to accurately predict the corrosion rate, the findings demonstrated the value of ML’s regression and data mining capabilities for corrosion data analysis. With the research gap in material selection strategies, proposed research will be critical to advancing data analytics approach exploiting their properties for high temperature corrosion applications.Graphical

A 2D VMD video image processing-based transfer learning approach for the detection and estimation of biofouling in tidal stream turbines

Harnessing the power of tidal streams is a sustainable way of exploiting renewable marine energy resources. It involves installing tidal stream turbines underwater to harness the energy. Nevertheless, these turbines are prone to the accumulation of biofouling, which significantly reduces their energy output and operational efficiency. It is therefore crucial to implement a condition-based monitoring system to detect biofouling promptly and ensure the continuous operation of a tidal stream turbine. In this context, this paper presents a data-centric approach that uses model submerged tidal stream turbine video images to detect and quantify biofouling. The relevance of a two-dimensional variational mode decomposition approach is investigated to extract relevant information from the potentially noisy collected images. While generative adversarial networks are used to address the data imbalance problem, a convolutional neural network is adopted to detect and assess the extent of biofouling. The performance of the proposed approach is assessed and validated using two experimental datasets obtained from the tidal stream turbine platforms of the Shanghai Maritime University and the Lehigh University.

High-quality expert annotations enhance artificial intelligence model accuracy for osteosarcoma X-ray diagnosis.

Primary malignant bone tumors, such as osteosarcoma, significantly affect the pediatric and young adult populations, necessitating early diagnosis for effective treatment. This study developed a high-performance artificial intelligence (AI) model to detect osteosarcoma from X-ray images using highly accurate annotated data to improve diagnostic accuracy at initial consultations. Traditional models trained on unannotated data have shown limited success, with sensitivities of approximately 60%-70%. In contrast, our model used a data-centric approach with annotations from an experienced oncologist, achieving a sensitivity of 95.52%, specificity of 96.21%, and an area under the curve of 0.989. The model was trained using 468 X-ray images from 31 osteosarcoma cases and 378 normal knee images with a strategy to maximize diversity in the training and validation sets. It was evaluated using an independent dataset of 268 osteosarcoma and 554 normal knee images to ensure generalizability. By applying the U-net architecture and advanced image processing techniques such as renormalization and affine transformations, our AI model outperforms existing models, reducing missed diagnoses and enhancing patient outcomes by facilitating earlier treatment. This study highlights the importance of high-quality training data and advocates a shift towards data-centric AI development in medical imaging. These insights can be extended to other rare cancers and diseases, underscoring the potential of AI in transforming diagnostic processes in oncology. The integration of this AI model into clinical workflows could support physicians in early osteosarcoma detection, thereby improving diagnostic accuracy and patient care.

Implementation of an Atrioventricular Valve Intervention Registry: Comparative Study of REDCap vs. CDR-Based openEHR Registry.

This comparative study examines the transition from isolated registries to a consolidated data-centric approach at University Hospital Schleswig-Holstein, focusing on migrating the Atrioventricular Valve Intervention Registry (AVIR) from REDCap to a Medical Data Integration Center based openEHR registry. Through qualitative analysis, we identify key disparities and strategic decisions guiding this transition. While REDCap has historical utility, its limitations in automated data integration and traceability highlight the advantages of a data-centric approach, which include streamlined data (integration) management at a single-point-of-truth based on e.g., centralized consent management. Our findings lay the groundwork for the AVIR project and a proof-of-concept data-centric registry, reflecting a broader industry trend towards data-centric healthcare initiatives.

Saliency-based video summarization for face anti-spoofing

With the growing availability of databases for face presentation attack detection, researchers are increasingly focusing on video-based face anti-spoofing methods that involve hundreds to thousands of images for training the models. However, there is currently no clear consensus on the optimal number of frames in a video to improve face spoofing detection. Inspired by the visual saliency theory, we present a video summarization method for face anti-spoofing detection that aims to enhance the performance and efficiency of deep learning models by leveraging visual saliency. In particular, saliency information is extracted from the differences between the Laplacian and Wiener filter outputs of the source images, enabling the identification of the most visually salient regions within each frame. Subsequently, the source images are decomposed into base and detail images, enhancing the representation of the most important information. Weighting maps are then computed based on the saliency information, indicating the importance of each pixel in the image. By linearly combining the base and detail images using the weighting maps, the method fuses the source images to create a single representative image that summarizes the entire video. The key contribution of the proposed method lies in demonstrating how visual saliency can be used as a data-centric approach to improve the performance and efficiency for face presentation attack detection. By focusing on the most salient images or regions within the images, a more representative and diverse training set can be created, potentially leading to more effective models. To validate the method’s effectiveness, a simple CNN–RNN deep learning architecture was used, and the experimental results showcased state-of-the-art performance on four challenging face anti-spoofing datasets.

Enhancing COVID-19 forecasting precision through the integration of compartmental models, machine learning and variants

Predicting epidemic evolution is essential for making informed decisions and guiding the implementation of necessary countermeasures. Computational models are vital tools that provide insights into illness progression and enable early detection, proactive intervention, and targeted preventive measures. This paper introduces Sybil, a framework that integrates machine learning and variant-aware compartmental models, leveraging a fusion of data-centric and analytic methodologies. To validate and evaluate Sybil’s forecasts, we employed COVID-19 data from several European and U.S. states. The dataset included the number of new and recovered cases, fatalities, and variant presence over time. We evaluate the forecasting precision of Sybil in periods in which there is a change in the trend of the pandemic evolution or a new variant appears. Results demonstrate that Sybil outperforms conventional data-centric approaches, being able to forecast accurately the changes in the trend, the magnitude of these changes, and the future prevalence of new variants.

An ensemble approach for enhanced Day-Ahead price forecasting in electricity markets

Electricity price forecasting (EPF) is a crucial aspect of daily trading operations, enabling market participants to make informed decisions regarding their bidding strategies. This paper explores a day-ahead price forecasting system that harnesses the potential of multiple machine learning (ML) models and their synergistic integration. This approach is designed to capitalize on the strengths of these models while also accounting for the unique characteristics of energy markets. For this purpose, several aggregation models were developed combining the predictions from ML models based on historical evaluations of their performance. The main objective of this approach is to enhance prediction accuracy by shifting the focus away from rigid model selection and instead prioritizing a data-centric approach, by focusing on data quality rather than rigid model selection. As a case study, the German energy market was examined due to its pivotal role within the EU system. The experimental results from this study provide valuable insights into the proposed system’s effectiveness and functionality.

Lean-and-Green Fractional Factorial Screening of 3D-Printed ABS Mechanical Properties Using a Gibbs Sampler and a Neutrosophic Profiler

The use of acrylonitrile butadiene styrene (ABS) in additive manufacturing applications constitutes an elucidating example of a promising match of a sustainable material to a sustainable production process. Lean-and-green datacentric-based techniques may enhance the sustainability of product-making and process-improvement efforts. The mechanical properties—the yield strength and the ultimate compression strength—of 3D-printed ABS product specimens are profiled by considering as many as eleven controlling factors at the process/product design stage. A fractional-factorial trial planner is used to sustainably suppress by three orders of magnitude the experimental needs for materials, machine time, and work hours. A Gibbs sampler and a neutrosophic profiler are employed to treat the complex production process by taking into account potential data uncertainty complications due to multiple distributions and indeterminacy issues due to inconsistencies owing to mechanical testing conditions. The small-data multifactorial screening outcomes appeared to steadily converge to three factors (the layer height, the infill pattern angle, and the outline overlap) with a couple of extra factors (the number of top/bottom layers and the infill density) to supplement the linear modeling effort and provide adequate predictions for maximizing the responses of the two examined mechanical properties. The performance of the optimal 3D-printed ABS specimens exhibited sustainably acceptable discrepancies, which were estimated at 3.5% for the confirmed mean yield strength of 51.70 MPa and at 5.5% for the confirmed mean ultimate compression strength of 53.58 MPa. The verified predictors that were optimally determined from this study were (1) the layer thickness—set at 0.1 mm; (2) the infill angle—set at 0°; (3) the outline overlap—set at 80%; (4) the number of top/bottom layers—set at 5; and (5) the infill density—set at 100%. The multifactorial datacentric approach composed of a fractional-factorial trial planner, a Gibbs sampler, and a neutrosophic profiler may be further tested on more intricate materials and composites while introducing additional product/process characteristics.

Uncertainty quantification by direct propagation of shallow ensembles

Statistical learning algorithms provide a generally-applicable framework to sidestep time-consuming experiments, or accurate physics-based modeling, but they introduce a further source of error on top of the intrinsic limitations of the experimental or theoretical setup. Uncertainty estimation is essential to quantify this error, and to make application of data-centric approaches more trustworthy. To ensure that uncertainty quantification is used widely, one should aim for algorithms that are accurate, but also easy to implement and apply. In particular, including uncertainty quantification on top of an existing architecture should be straightforward, and add minimal computational overhead. Furthermore, it should be easy to manipulate or combine multiple machine-learning predictions, propagating uncertainty over further modeling steps. We compare several well-established uncertainty quantification frameworks against these requirements, and propose a practical approach, which we dub direct propagation of shallow ensembles, that provides a good compromise between ease of use and accuracy. We present benchmarks for generic datasets, and an in-depth study of applications to the field of atomistic machine learning for chemistry and materials. These examples underscore the importance of using a formulation that allows propagating errors without making strong assumptions on the correlations between different predictions of the model.

3D GeoRemediation: A Digital Hydrogeophysical–Chemical Clone and Virtual Hydraulic Barrier with Groundwater Circulation Wells (GCWs) for Groundwater Remediation

Identification of contamination sources and delineation of plumes in the geological environment stand as pivotal elements in reconstructing the conceptual site model (CSM) and devising remediation strategies tailored to specific physicochemical traits. This study endeavors to showcase the capabilities of a 3D digital interface, seamlessly integrating multi-source data, to elucidate site-specific contamination dynamics and steer the implementation of remediation strategies harmoniously aligned with the ethos of remediation geology. In a site historically marred by chlorinated solvent contamination, the digitization of stratigraphic, piezometric, chemical, and membrane interface probe (MIP) data underpins geomodeling endeavors and yields a meticulously crafted, data-driven CSM. The hydrogeochemical and hydrogeophysical data were interpolated to build a volumetric, digital 3D model illustrating data-driven elements. The comprehensive 3D clone adeptly delineates secondary contamination sources and renders visible the contamination plume within a georeferenced framework, mirroring the nuanced interplay of stratigraphic nuances and groundwater path. A data-centric approach to modeling facilitates the design of the first hydraulic virtual barrier leveraging groundwater circulation well (GCW) technology, its geometry finely attuned to intercept the contamination plume originating from source dissolution and aligning with preferential groundwater flow trajectories. Conventional hydrochemical monitoring and multilevel sampling substantiate the discernible reduction in chlorinated solvent concentrations across various depths within the aquifer horizon, affirming the efficacy of GCWs in their virtual barrier configuration. The findings highlight the effectiveness and limited groundwater consumption of the virtual barrier compared to the on-site pump-and-stock system. This research underscores the potency of a multi-faceted evidence-driven puzzle in conceptualizing contamination mechanisms within the geological milieu, thereby fostering the application of cutting-edge, effective, and sustainable remediation strategies.

PalmSecMatch: A data-centric template protection method for palmprint recognition

While existing palmprint recognition researches aim to improve accuracy in various situations, they often overlook the security implications. This paper delves into template protection in palmprint recognition. The existing template protection methods usually cannot strike a well balance between security, accuracy and usability, which reduces the applicability of the algorithms. In this work, a data-centric approach for palmprint template protection is proposed, called PalmSecMatch. Our solution extracts the key from plaintext data. It extremely reduces the dependency on third-party or independent key generation algorithms. The backbone of PalmSecMatch consists of key data extraction and encryption, order shuffling of the raw vectors, hashing code generation, shuffling basis and hashing code fading. PalmSecMatch subtly exploits the fact that biometric data are random variables and benefits from its data-centric nature. PalmSecMatch allows the same plaintext features to be encrypted into highly different ciphertexts, which greatly ensures security. At the same time, the application of data fading strategy makes it extremely difficult for an attacker to distinguish the user data from the auxiliary data. The security analysis shows that PalmSecMatch satisfies the requirements of ISO/IEC 24745. Adequate experiments on two public palmprint databases validate the effectiveness of the proposed method.

Novel Insights in Soil Mechanics: Integrating Experimental Investigation with Machine Learning for Unconfined Compression Parameter Prediction of Expansive Soil

This paper presents a novel application of machine learning models to clarify the intricate behaviors of expansive soils, focusing on the impact of sand content, saturation level, and dry density. Departing from conventional methods, this research utilizes a data-centric approach, employing a suite of sophisticated machine learning models to predict soil properties with remarkable precision. The inclusion of a 30% sand mixture is identified as a critical threshold for optimizing soil strength and stiffness, a finding that underscores the transformative potential of sand amendment in soil engineering. In a significant advancement, the study benchmarks the predictive power of several models including extreme gradient boosting (XGBoost), gradient boosting regression (GBR), random forest regression (RFR), decision tree regression (DTR), support vector regression (SVR), symbolic regression (SR), and artificial neural networks (ANNs and proposed ANN-GMDH). Symbolic regression equations have been developed to predict the elasticity modulus and unconfined compressive strength of the investigated expansive soil. Despite the complex behaviors of expansive soil, the trained models allow for optimally predicting the values of unconfined compressive parameters. As a result, this paper provides for the first time a reliable and simply applicable approach for estimating the unconfined compressive parameters of expansive soils. The proposed ANN-GMDH model emerges as the pre-eminent model, demonstrating exceptional accuracy with the best metrics. These results not only highlight the ANN’s superior performance but also mark this study as a groundbreaking endeavor in the application of machine learning to soil behavior prediction, setting a new benchmark in the field.

Large-scale Probabilistic Forecasting of Consumer Engagement of CPG Products using Heterogeneous Web Data

Consumer Packaged Goods (CPG) play a pivotal role in customer-centric industries. Understanding the distinctive features of such products and how customers engage with them is essential for CPG manufacturers to create customer-winning products. In this study, we explore an innovative data-centric approach to forecast customer engagement for CPG products by monitoring their digital evolution, particularly in the Snacks category in the USA. Traditional methods for consumer analysis such as surveys, focus groups are time consuming, costly and ineffective with limited scope. They also lack a forecasting component, making it difficult for CPG companies to make forward-looking decisions. However, with the emergence of big data, we can leverage user generated public data from social media, web search, e-commerce platforms to estimate consumer engagement and make long-term forecasts. To achieve this, we propose a systematic approach to accumulate and prepare large datasets from heterogeneous web sources for CPG products. We then use state-of-the-art deep learning based time series forecasting models to efficiently train and predict consumer engagement for the next 12 months, benchmarking their computational efficiency and forecasting performance. Our findings indicate that the DeepAR model outperforms all other models, with the lowest NRMSE (=0.378), RMSE (=14.848) MSE (=220.457), MASE (=0.871), and sMAPE (=0.306) values. Furthermore, we demonstrate methods for computing single forecasting points and prediction intervals using the forecasted sample distribution from the probabilistic models. The proposed approach will provide CPG businesses with valuable insights to make informed decisions about product development, marketing strategies, and supply chain management.

Reliability evaluation of individual predictions: a data-centric approach

Reliability evaluation of individual predictions: a data-centric approach

Leveraging Limited Experimental Data with Machine Learning: Differentiating a Methyl from an Ethyl Group in the Corey-Bakshi-Shibata Reduction.

We present a case study on how to improve an existing metal-free catalyst for a particularly difficult reaction, namely, the Corey-Bakshi-Shibata (CBS) reduction of butanone, which constitutes the classic and prototypical challenge of being able to differentiate a methyl from an ethyl group. As there are no known strategies on how to address this challenge, we leveraged the power of machine learning by constructing a realistic (for a typical laboratory) small, albeit high-quality, data set of about 100 reactions (run in triplicate) that we used to train a model in combination with a key-intermediate graph (of substrate and catalyst) to predict the differences in Gibbs activation energies ΔΔG‡ of the enantiomeric reaction paths. With the help of this model, we were able to select and subsequently screen a small selection of catalysts and increase the selectivity for the CBS reduction of butanone to 80% enantiomeric excess (ee), the highest possible value achieved to date for this substrate with a metal-free catalyst, thereby also exceeding the best available enzymatic systems (64% ee) and the selectivity with Corey's original catalyst (60% ee). This translates into a >50% improvement in relative ΔG‡ from 0.9 to 1.4 kcal mol-1. We underscore the transformative potential of machine learning in accelerating catalyst design because we rely on a manageable small data set and a key-intermediate graph representing a combination of catalyst and substrate graphs in lieu of a transition-state model. Our results highlight the synergy of synthetic chemistry and data-centric approaches and provide a blueprint for future catalyst optimization.