Prediction of the Unconfined Compressive Strength of a One-Part Geopolymer-Stabilized Soil Using Deep Learning Methods with Combined Real and Synthetic Data DOI Creative Commons
Qinyi Chen,

Guo Hu,

Jun Wu

и другие.

Buildings, Год журнала: 2024, Номер 14(9), С. 2894 - 2894

Опубликована: Сен. 13, 2024

This study focused on exploring the utilization of a one-part geopolymer (OPG) as sustainable alternative binder to ordinary Portland cement (OPC) in soil stabilization, offering significant environmental advantages. The unconfined compressive strength (UCS) was key index for evaluating efficacy OPG traditionally demanding substantial resources terms cost and time. In this research, four distinct deep learning (DL) models (Artificial Neural Network [ANN], Backpropagation [BPNN], Convolutional [CNN], Long Short-Term Memory [LSTM]) were employed predict UCS OPG-stabilized soft clay, providing more efficient precise methodology. Among these models, CNN exhibited highest performance (MAE = 0.022, R2 0.9938), followed by LSTM 0.0274, 0.9924) BPNN 0.0272, 0.9921). Wasserstein Generative Adversarial (WGAN) further utilized generate additional synthetic samples expanding training dataset. incorporation generated WGAN into set DL led improved performance. When number achieved 200, WGAN-CNN model provided most accurate results, with an value 0.9978 MAE 0.9978. Furthermore, assess reliability gain insights influence input variables predicted outcomes, interpretable Machine Learning techniques, including sensitivity analysis, Shapley Additive Explanation (SHAP), 1D Partial Dependence Plot (PDP) analyzing interpreting models. research illuminates new aspects application real data properties soil, contributing saving time cost.

Язык: Английский

Security Considerations in Generative AI for Web Applications DOI
Siva Raja Sindiramutty,

Krishna Raj V. Prabagaran,

N. Z. Jhanjhi

и другие.

Advances in information security, privacy, and ethics book series, Год журнала: 2024, Номер unknown, С. 281 - 332

Опубликована: Июль 26, 2024

Protecting AI in web applications is necessary. This domain a composite of technology and huge scope with good prospects immense difficulties. chapter covers the landscape security issues advancing generative techniques for integration into development frameworks. The initial section on development—a conversation subtleties AI-based methods. In literal stance, offers 13 ways to approach it. Among threats are those that introduce related deployments, which illustrate why it vital defenders infrastructure owners implement mitigation measures proactively. pertains privacy data lessons securing preventing vulnerability. explores attacks, model poisoning, bias issues, defence mechanisms, long-term strategies. Additionally, Service A promotes transparency, explainability, compliance applicable laws while structuring methodology deployment methods/operation. text outlines how respond recover from incidents as provides response frameworks everyone involved managing breaches. Finally, addresses trends, possible threats, learned real-world case studies. order contribute addressing these research needs, this sheds light considerations associated suggests recommendations can help researchers, practitioners, policymakers enhance posture popular advancements used generating applications.

Язык: Английский

Процитировано

2
Restructuring the Landscape of Generative AI Research DOI
Salaheldin Edam

Advances in educational technologies and instructional design book series, Год журнала: 2024, Номер unknown, С. 287 - 334

Опубликована: Сен. 20, 2024

This Chapter delves into the impact of generative AI on academic research and publishing, discussing various architectures such as Mixture Experts (MoE), Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), Pre-trained Transformers (GPT). The explores increase AI-centered preprints, their effects peer review, ethical considerations linked to them. peer-review system's integrity is under examination, focusing challenges related AI, misuse, redefining plagiarism. chapter potential tools improve review processes stresses importance institutions creating frameworks for utilization. article concludes by evaluating advantages drawbacks in research, with goal presenting a fair viewpoint its revolutionary capabilities while upholding principles.

Язык: Английский

Процитировано

1
Deep learning for safety risk management in modular construction: Status, strengths, challenges, and future directions DOI
Yin Junjia, Aidi Hizami Alias, Nuzul Azam Haron

и другие.

Automation in Construction, Год журнала: 2024, Номер 169, С. 105894 - 105894

Опубликована: Ноя. 28, 2024

Язык: Английский

Процитировано

1
Prediction of the Unconfined Compressive Strength of a One-Part Geopolymer-Stabilized Soil Using Deep Learning Methods with Combined Real and Synthetic Data DOI Creative Commons
Qinyi Chen,

Guo Hu,

Jun Wu

и другие.

Buildings, Год журнала: 2024, Номер 14(9), С. 2894 - 2894

Опубликована: Сен. 13, 2024

This study focused on exploring the utilization of a one-part geopolymer (OPG) as sustainable alternative binder to ordinary Portland cement (OPC) in soil stabilization, offering significant environmental advantages. The unconfined compressive strength (UCS) was key index for evaluating efficacy OPG traditionally demanding substantial resources terms cost and time. In this research, four distinct deep learning (DL) models (Artificial Neural Network [ANN], Backpropagation [BPNN], Convolutional [CNN], Long Short-Term Memory [LSTM]) were employed predict UCS OPG-stabilized soft clay, providing more efficient precise methodology. Among these models, CNN exhibited highest performance (MAE = 0.022, R2 0.9938), followed by LSTM 0.0274, 0.9924) BPNN 0.0272, 0.9921). Wasserstein Generative Adversarial (WGAN) further utilized generate additional synthetic samples expanding training dataset. incorporation generated WGAN into set DL led improved performance. When number achieved 200, WGAN-CNN model provided most accurate results, with an value 0.9978 MAE 0.9978. Furthermore, assess reliability gain insights influence input variables predicted outcomes, interpretable Machine Learning techniques, including sensitivity analysis, Shapley Additive Explanation (SHAP), 1D Partial Dependence Plot (PDP) analyzing interpreting models. research illuminates new aspects application real data properties soil, contributing saving time cost.

Язык: Английский

Процитировано

0