Method for Rail Surface Defect Detection Based on Neural Network Architecture Search DOI
Yongzhi Min,

Qinglong Jing,

Yaxing Li

et al.

Measurement Science and Technology, Journal Year: 2024, Volume and Issue: 36(1), P. 016027 - 016027

Published: Nov. 8, 2024

Abstract This study addresses the inherent limitations of implementing neural network architecture search algorithms for rail surface defect detection, including low efficiency and oversight edge features on surface. A sophisticated multi-level framework is proposed that integrates emphasizes features. The utilizes Z-Score normalization method to quantify concern samples, combined with an Edge-Loss function enhance feature recognition capabilities. Furthermore, acknowledging sensitivity spatial resolution changes, a space meticulously designed. In cell-level space, combining partial channel sampling operation pruning employed model regularization. network-level optimal paths change are established, allowing screening aggregation at various levels facilitate adaptive extraction multi-scale Experimental outcomes indicate this significantly reduces computational resource usage by approximately 75% increases mIOU 2.6% relative traditional methods. Moreover, it demonstrates robust capability in accurately recognizing defective edges surfaces, thereby substantiating method’s effectiveness.

Language: Английский

Advancing cancer diagnosis and prognostication through deep learning mastery in breast, colon, and lung histopathology with ResoMergeNet DOI
Chukwuebuka Joseph Ejiyi, Zhen Qin, Victor Kwaku Agbesi

et al.

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 185, P. 109494 - 109494

Published: Dec. 4, 2024

Language: Английский

Citations

6
PKMT-Net: A pathological knowledge-inspired multi-scale transformer network for subtype prediction of lung cancer using histopathological images DOI

Zhilei Zhao,

Shuli Guo,

Lina Han

et al.

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107742 - 107742

Published: Feb. 21, 2025

Language: Английский

Citations

0
A new multi-objective hyperparameter optimization algorithm for COVID-19 detection from x-ray images DOI Creative Commons
Burak Gülmez

Soft Computing, Journal Year: 2024, Volume and Issue: 28(19), P. 11601 - 11617

Published: July 23, 2024

Abstract The coronavirus occurred in Wuhan (China) first and it was declared a global pandemic. To detect X-ray images can be used. Convolutional neural networks (CNNs) are used commonly to illness from images. There lots of different alternative deep CNN models or architectures. find the best architecture, hyper-parameter optimization In this study, problem is modeled as multi-objective (MOO) problem. Objective functions multi-class cross entropy, error ratio, complexity network. For solutions objective functions, made by NSGA-III, NSGA-II, R-NSGA-II, SMS-EMOA, MOEA/D, proposed Swarm Genetic Algorithms (SGA). SGA swarm-based algorithm with cross-over process. All six algorithms run give Pareto optimal solution sets. When figures obtained analyzed hypervolume values compared, outperforms MOEA/D algorithms. It concluded that better than others for COVID-19 detection Also, sensitivity analysis has been understand effect number parameters on model success.

Language: Английский

Citations

3
Enhancing infectious disease prediction model selection with multi-objective optimization: an empirical study DOI Creative Commons

Deren Xu,

Weng Howe Chan, Habibollah Haron

et al.

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e2217 - e2217

Published: July 29, 2024

As the pandemic continues to pose challenges global public health, developing effective predictive models has become an urgent research topic. This study aims explore application of multi-objective optimization methods in selecting infectious disease prediction and evaluate their impact on improving accuracy, generalizability, computational efficiency. In this study, NSGA-II algorithm was used compare selected by with those traditional single-objective optimization. The results indicate that decision tree (DT) extreme gradient boosting regressor (XGBoost) through outperform other terms Compared ridge regression model methods, XGBoost demonstrate significantly lower root mean square error (RMSE) real datasets. finding highlights potential advantages balancing multiple evaluation metrics. However, study's limitations suggest future directions, including improvements, expanded metrics, use more diverse conclusions emphasize theoretical practical significance health support systems, indicating wide-ranging applications models.

Language: Английский

Citations

1
Pitfalls in Metaheuristics Solving Stoichiometric-Based Optimization Models for Metabolic Networks DOI Creative Commons
Mónica Fabiola Briones-Báez, Luciano Aguilera‐Vázquez, Nelson Rangel-Valdez

et al.

Algorithms, Journal Year: 2024, Volume and Issue: 17(8), P. 336 - 336

Published: Aug. 1, 2024

Flux Balance Analysis (FBA) is a constraint-based method that commonly used to guide metabolites through restricting pathways often involve conditions such as anaplerotic cycles like Calvin, reversible or irreversible reactions, and nodes where metabolic branch. The can identify the best for one course but fails when dealing with of multiple interest. Recent studies on metabolism consider it more natural optimize several simultaneously rather than just one; moreover, they point out use metaheuristics an attractive alternative extends FBA tackle objectives. However, literature also warns techniques must not be wild. Instead, subject careful fine-tuning selection processes achieve desired results. This work analyses impact quality built using NSGAII MOEA/D algorithms novel optimization models; conducts study two case studies, pigment biosynthesis node in glutamate microalgae Chlorella vulgaris, under three culture (autotrophic, heterotrophic, mixotrophic) while optimizing intermediaries independent objective functions simultaneously. results show varying performances between MOEA/D, demonstrating model greatly affect predicted phenotypes.

Language: Английский

Citations

1
Method for Rail Surface Defect Detection Based on Neural Network Architecture Search DOI
Yongzhi Min,

Qinglong Jing,

Yaxing Li

et al.

Measurement Science and Technology, Journal Year: 2024, Volume and Issue: 36(1), P. 016027 - 016027

Published: Nov. 8, 2024

Abstract This study addresses the inherent limitations of implementing neural network architecture search algorithms for rail surface defect detection, including low efficiency and oversight edge features on surface. A sophisticated multi-level framework is proposed that integrates emphasizes features. The utilizes Z-Score normalization method to quantify concern samples, combined with an Edge-Loss function enhance feature recognition capabilities. Furthermore, acknowledging sensitivity spatial resolution changes, a space meticulously designed. In cell-level space, combining partial channel sampling operation pruning employed model regularization. network-level optimal paths change are established, allowing screening aggregation at various levels facilitate adaptive extraction multi-scale Experimental outcomes indicate this significantly reduces computational resource usage by approximately 75% increases mIOU 2.6% relative traditional methods. Moreover, it demonstrates robust capability in accurately recognizing defective edges surfaces, thereby substantiating method’s effectiveness.

Language: Английский

Citations

0