article · Scientific Reports
Deep learning research has traditionally concentrated on developing new algorithms and model architectures while keeping training datasets fixed. An alternative data-centric approach focuses instead on enhancing dataset quality. This research assesses a data-centric methodology that combines data augmentation, multi-stage hashing to remove duplicate records, and confident learning to identify and correct mislabelled instances. Benchmarking experiments were conducted using the ResNet-18 neural network architecture across three standard image datasets: MNIST, Fashion MNIST, and CIFAR-10. Across these evaluations, the data-centric strategy consistently improved classification performance over the conventional model-centric strategy by a relative margin of at least 3%. Systematic dataset curation can significantly elevate model accuracy, showing clear promise for domains that rely heavily on reliable data, including healthcare, finance, education, and entertainment.
Artificial intelligence systems often underperform when trained on noisy, duplicated, or mislabelled information. By demonstrating that systematic data refinement delivers a measurable boost in model accuracy, this work highlights how organisations can achieve superior predictive performance through disciplined dataset management rather than relying solely on larger, more resource-intensive computational models.
The abstract highlights potential applications in healthcare, finance, education, and entertainment, where dataset reliability is critical. The primary prospective users are machine learning engineers and data teams. Because the methodology was evaluated on standard benchmark image sets using ResNet-18, the work represents early-stage experimental research that has not yet been demonstrated in commercial environments or on domain-specific enterprise data.
AI-generated from the published abstract. Always read the original work before citing.
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.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1038/s41598-024-73643-x
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
