Advanced machine learning models for accurate and efficient anomaly detection in scanned visual data
-
Mangala Shetty
Abstract
Combination of inputs as multimodal signals and their dynamic nature as modern data call for more sophisticated approaches to anomaly detectionanomaly detection. The literature review of this book chapter focuses on using a combination of supervised, unsupervised, and reinforcement learning to overcome the drawbacks of conventional approaches. Some of them are ensemble methods, basically the hybrid deep learning framework and multimodal deep learning frameworks involving transfer learning and also features fusion to tackle the anomaly detection problems in different fields. Additional attention is paid to the integration of the algorithms including convolutional neural networks, recurrent neural networks, generative adversarial networks, and variational autoencoders in the loops for improvement of the feature extracting algorithm and its reliability. Further, for enhanced anomaly detection, policies that are learned are described, which include policy gradient methods and deep Q-learningdeep Q-learning. Multimodal anomaly detection is then discussed with further analysis methods that utilize data originating from dissimilar sources of information and context such as image data, sensor data, and metadata. Many issues are discussed within the chapter, such as data imbalance, cross-modal transfer, and computational limitations, whereas solutions are presented in terms of ensembling and hybrid approaches. Implementations in many fields including healthcare, production, and cyber-security are also reviewed to show different uses of hybrid models. Addressing the findings further, this chapter ends with the insights on directions for future work: scalable algorithms and self-supervised learning, as well as the importance of explainable AI in the context of anomaly detection.
Abstract
Combination of inputs as multimodal signals and their dynamic nature as modern data call for more sophisticated approaches to anomaly detectionanomaly detection. The literature review of this book chapter focuses on using a combination of supervised, unsupervised, and reinforcement learning to overcome the drawbacks of conventional approaches. Some of them are ensemble methods, basically the hybrid deep learning framework and multimodal deep learning frameworks involving transfer learning and also features fusion to tackle the anomaly detection problems in different fields. Additional attention is paid to the integration of the algorithms including convolutional neural networks, recurrent neural networks, generative adversarial networks, and variational autoencoders in the loops for improvement of the feature extracting algorithm and its reliability. Further, for enhanced anomaly detection, policies that are learned are described, which include policy gradient methods and deep Q-learningdeep Q-learning. Multimodal anomaly detection is then discussed with further analysis methods that utilize data originating from dissimilar sources of information and context such as image data, sensor data, and metadata. Many issues are discussed within the chapter, such as data imbalance, cross-modal transfer, and computational limitations, whereas solutions are presented in terms of ensembling and hybrid approaches. Implementations in many fields including healthcare, production, and cyber-security are also reviewed to show different uses of hybrid models. Addressing the findings further, this chapter ends with the insights on directions for future work: scalable algorithms and self-supervised learning, as well as the importance of explainable AI in the context of anomaly detection.
Kapitel in diesem Buch
- Frontmatter I
- Contents V
- List of contributors VII
- Deep learning in computer vision 1
- Deep learning for medical image segmentation 51
- Deep learning for image segmentation 107
- Machine learning algorithm for medical image processing 155
- Machine learning models for predicting anomaly in scanned images 215
- Advanced machine learning models for accurate and efficient anomaly detection in scanned visual data 263
- AI-enhanced diagnostic materials improving sensitivity for disease detection and diagnostics 311
- Machine learning approaches for optimizing the synthesis and functionalization of quantum dots for medical imaging 353
- Machine learning application in tissue engineering: scaffold design 407
- Machine learning approaches to improve electrospun nanofibers’ performance and properties for medical applications 441
- Predictive machine learning models for assessing the long-term stability of biodegradable scaffolds 483
- Customization of medical implants using 3D printing 523
- Index 559
- De Gruyter Series in Advanced Mechanical Engineering
Kapitel in diesem Buch
- Frontmatter I
- Contents V
- List of contributors VII
- Deep learning in computer vision 1
- Deep learning for medical image segmentation 51
- Deep learning for image segmentation 107
- Machine learning algorithm for medical image processing 155
- Machine learning models for predicting anomaly in scanned images 215
- Advanced machine learning models for accurate and efficient anomaly detection in scanned visual data 263
- AI-enhanced diagnostic materials improving sensitivity for disease detection and diagnostics 311
- Machine learning approaches for optimizing the synthesis and functionalization of quantum dots for medical imaging 353
- Machine learning application in tissue engineering: scaffold design 407
- Machine learning approaches to improve electrospun nanofibers’ performance and properties for medical applications 441
- Predictive machine learning models for assessing the long-term stability of biodegradable scaffolds 483
- Customization of medical implants using 3D printing 523
- Index 559
- De Gruyter Series in Advanced Mechanical Engineering
