A survey on deep learning tools dealing with data scarcity: definitions, challenges, solutions, tips, and applications

Journal Of Big Data, Journal Year: 2023, Volume and Issue: 10(1)

Published: April 14, 2023

Abstract Data scarcity is a major challenge when training deep learning (DL) models. DL demands large amount of data to achieve exceptional performance. Unfortunately, many applications have small or inadequate train frameworks. Usually, manual labeling needed provide labeled data, which typically involves human annotators with vast background knowledge. This annotation process costly, time-consuming, and error-prone. every framework fed by significant automatically learn representations. Ultimately, larger would generate better model its performance also application dependent. issue the main barrier for dismissing use DL. Having sufficient first step toward any successful trustworthy application. paper presents holistic survey on state-of-the-art techniques deal models overcome three challenges including small, imbalanced datasets, lack generalization. starts listing techniques. Next, types architectures are introduced. After that, solutions address listed, such as Transfer Learning (TL), Self-Supervised (SSL), Generative Adversarial Networks (GANs), Model Architecture (MA), Physics-Informed Neural Network (PINN), Deep Synthetic Minority Oversampling Technique (DeepSMOTE). Then, these were followed some related tips about acquisition prior purposes, well recommendations ensuring trustworthiness dataset. The ends list that suffer from scarcity, several alternatives proposed in order more each Electromagnetic Imaging (EMI), Civil Structural Health Monitoring, Medical imaging, Meteorology, Wireless Communications, Fluid Mechanics, Microelectromechanical system, Cybersecurity. To best authors’ knowledge, this review offers comprehensive overview strategies tackle

Language: Английский

---

PINN-FORM: A new physics-informed neural network for reliability analysis with partial differential equation

Computer Methods in Applied Mechanics and Engineering, Journal Year: 2023, Volume and Issue: 414, P. 116172 - 116172

Published: June 22, 2023

Language: Английский

---

A Review of Physics Informed Neural Networks for Multiscale Analysis and Inverse Problems

Multiscale Science and Engineering, Journal Year: 2024, Volume and Issue: 6(1), P. 1 - 11

Published: Feb. 13, 2024

Language: Английский

---

Estimation of load for tunnel lining in elastic soil using physics‐informed neural network

Computer-Aided Civil and Infrastructure Engineering, Journal Year: 2024, Volume and Issue: 39(17), P. 2701 - 2718

Published: April 10, 2024

Abstract A reverse calculation method termed soil and lining physics‐informed neural network (SL‐PINN) is proposed for the estimation of load tunnel in elastic based on radial displacement measurements lining. To achieve efficient accurate calculations, framework SL‐PINN specially designed to consider respective characteristics surrounding multistep training meshless established promote efficiency. The involves increasing number collocation points each step while decreasing learning rate after scaling SL‐PINN. feasibility verified by numerical simulation data field data. Compared other inverse methods, has lower precision requirements measurement instrument with same level accuracy.

Language: Английский

---

Solutions to Two‐ and Three‐Dimensional Incompressible Flow Fields Leveraging a Physics‐Informed Deep Learning Framework and Kolmogorov–Arnold Networks

International Journal for Numerical Methods in Fluids, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 2, 2025

ABSTRACT Physics‐informed neural network (PINN) has become a potential technology for fluid dynamics simulations, but traditional PINN low accuracy in simulating incompressible flows, and these problems can lead to not converging. This paper proposes physics‐informed method (KA‐PINN) based on the Kolmogorov–Arnold Neural (KAN) structure. It is used solve two‐dimensional three‐dimensional problems. The flow field reconstructed predicted Kovasznay Beltrami flow. results show that prediction of KA‐PINN improved by about 5 times two dimensions 2 three compared with fully connected structure PINN. Meanwhile, number parameters reduced 8 10 times. research only verify application also demonstrate feasibility KAN improving ability predict fields. study reduce dependence numerical methods solving

Language: Английский

---

A review on the applications of artificial neural network techniques for accelerating finite element analysis in the civil engineering domain

Computers & Structures, Journal Year: 2025, Volume and Issue: 310, P. 107698 - 107698

Published: Feb. 26, 2025

Language: Английский

---

A complete Physics-Informed Neural Network-based framework for structural topology optimization

Computer Methods in Applied Mechanics and Engineering, Journal Year: 2023, Volume and Issue: 417, P. 116401 - 116401

Published: Sept. 9, 2023

Physics-Informed Neural Networks (PINNs) have recently gained increasing attention in the field of topology optimization. The fusion deep learning and optimization has emerged as a prominent area insightful research, where minimization loss function neural networks can be comparable to objective Inspired by concepts PINNs, this paper proposes novel framework, 'Complete Network-based Topology Optimization (CPINNTO)', address various challenges optimization, particularly related structural key innovation proposed framework lies introducing first complete machine-learning-based through integration two distinct PINNs. Herein, Deep Energy Method (DEM) PINN is implemented determine deformation state corresponding structures numerically. In addition, derivation with respect design variables replaced automatic differentiation sensitivity-analysis (S-PINN). feasibility potential CPINNTO been assessed several case studies while highlighting strengths limitations utilizing PINNs Subsequent findings indicate that achieve optimal topologies without labeled data nor FEA. numerical examples demonstrate capable stably obtaining for applications, including compliance problems, multi-constrained three-dimensional problems. Resulting designs exhibit favorable values obtained via density-based summary, opens up interesting possibilities

Language: Английский

---

Physics-informed radial basis network (PIRBN): A local approximating neural network for solving nonlinear partial differential equations

Computer Methods in Applied Mechanics and Engineering, Journal Year: 2023, Volume and Issue: 415, P. 116290 - 116290

Published: Aug. 3, 2023

Our recent intensive study has found that physics-informed neural networks (PINN) tend to be local approximators after training. This observation leads this novel radial basis network (PIRBN), which can maintain the property throughout entire training process. Compared deep networks, a PIRBN comprises of only one hidden layer and "activation" function. Under appropriate conditions, we demonstrated PIRBNs using gradient descendent methods converge Gaussian processes. Besides, studied dynamics via tangent kernel (NTK) theory. In addition, comprehensive investigations regarding initialisation strategies were conducted. Based on numerical examples, been more effective efficient than PINN in solving PDEs with high-frequency features ill-posed computational domains. Moreover, existing techniques, such as adaptive learning, decomposition different types loss functions, are applicable PIRBN. The programs regenerate all results at https://github.com/JinshuaiBai/PIRBN.

Language: Английский

---

Optimal parameters selection of back propagation algorithm in the feedforward neural network

Engineering Analysis with Boundary Elements, Journal Year: 2023, Volume and Issue: 151, P. 575 - 596

Published: March 30, 2023

Language: Английский

---

An Introduction to Programming Physics-Informed Neural Network-Based Computational Solid Mechanics

International Journal of Computational Methods, Journal Year: 2023, Volume and Issue: 20(10)

Published: May 5, 2023

Physics-informed neural network (PINN) has recently gained increasing interest in computational mechanics. This work extends the PINN to solid mechanics problems. Our focus will be on investigation of various formulation and programming techniques, when governing equations are implemented. Two prevailingly used physics-informed loss functions for PINN-based implemented examined. Numerical examples ranging from 1D 3D problems presented show performance The programs built via Python with TensorFlow library step-by-step explanations can extended more challenging applications. aims help researchers who interested solver have a clear insight into this emerging area. all numerical available at https://github.com/JinshuaiBai/PINN_Comp_Mech .

Language: Английский

---