Machine Learning with Noisy Labels
Definitions, Theory, Techniques and Solutions
- 1st Edition - February 23, 2024
- Imprint: Academic Press
- Author: Gustavo Carneiro
- Language: English
- Paperback ISBN:9 7 8 - 0 - 4 4 3 - 1 5 4 4 1 - 6
- eBook ISBN:9 7 8 - 0 - 4 4 3 - 1 5 4 4 2 - 3
Machine Learning and Noisy Labels: Definitions, Theory, Techniques and Solutions provides an ideal introduction to machine learning with noisy labels that is suitable for senior… Read more
Purchase options
Institutional subscription on ScienceDirect
Request a sales quoteMachine Learning and Noisy Labels: Definitions, Theory, Techniques and Solutions provides an ideal introduction to machine learning with noisy labels that is suitable for senior undergraduates, post graduate students, researchers and practitioners using, and researching, machine learning methods. Most of the modern machine learning models based on deep learning techniques depend on carefully curated and cleanly labeled training sets to be reliably trained and deployed. However, the expensive labeling process involved in the acquisition of such training sets limits the number and size of datasets available to build new models, slowing down progress in the field. This book defines the different types of label noise, introduces the theory behind the problem, presents the main techniques that enable the effective use of noisy-label training sets, and explains the most accurate methods.
- Shows how to design and reproduce regression, classification and segmentation models using large-scale noisy-label training sets
- Gives an understanding of the theory of, and motivation for, noisy-label learning
- Shows how to classify noisy-label learning methods into a set of core techniques
Senior undergraduates, graduate students and researchers in computer vision and medical imaging, machine learning, biomedical engineering, Radiologists
1: Problem Definition
1.1: Introduction and Motivation
1.2: Challenges
1.3: Survey Papers and Books on the Same Topic
1.4: Book Organisation
2: Noisy-label Problems and Datasets
2.1: Regression, Classification and Segmentation Problems
2.2: Closed Set Label Noise Problems
2.2.2: Symmetric
2.2.2: Asymmetric
2.2.3: Instance-dependent
2.3: Open Set Label Noise Problems
2.3.1: Symmetric
2.3.2: Asymmetric
2.3.3 Instance-dependent
2.4: Datasets
2.4.1 Computer Vision Datasets
2.4.1.1 Synthetic
2.4.1.2 Real
2.5 Medical Image Analysis Datasets
3: Theoretical Aspects of Noisy-label Learning
3.1: Bias Variance Decomposition
3.1.1: Regression
3.1.2: Classification
3.2: Taxonomy of Noisy Label
3.3: PAC Learning
4: Noisy-Label Learning Techniques
4.1: Loss Function
4.1.1: Label Noise Robust Loss
4.1.2: Loss Regularisation
4.1.3: Loss Reweighting
4.1.4: Loss Redesign
4.1.5: Loss Correction
4.2: Training Algorithms
4.2.1: Multi-stage Training
4.2.2: Meta-learning
4.2.3: Co-training
4.2.4: Self-training
4.2.5: Self-supervised Pre-training
4.2.6: Semi-supervised Learning
4.2.7: Expectation Maximization
4.3: Data Processing
4.3.1: Adversarial training
4.3.2: Label Cleaning
4.3.3: Sample Weighting
4.3.4: Sample Selection
4.3.5: Label Smoothing
4.4: Model Architecture
4.4.1: Adaptation Layer
4.4.2: Multiple Classifiers
4.4.3: Combining Generative and Discriminative models
5: Benchmarks, Methods, Results and Code
5.1 Closed Set Label Noise Problems
5.2 Open Set Label Noise Problems
5.3 Imbalanced Noisy-Label Problems
5.4 Imbalanced Noisy Multi-label Problems
5.5 Noisy-Label Segmentation Problems
5.6 How to Discover Future State-of-the-art Methods
6: Conclusion and Final Considerations
6.1 Conclusions
6.2 Future Work
1.1: Introduction and Motivation
1.2: Challenges
1.3: Survey Papers and Books on the Same Topic
1.4: Book Organisation
2: Noisy-label Problems and Datasets
2.1: Regression, Classification and Segmentation Problems
2.2: Closed Set Label Noise Problems
2.2.2: Symmetric
2.2.2: Asymmetric
2.2.3: Instance-dependent
2.3: Open Set Label Noise Problems
2.3.1: Symmetric
2.3.2: Asymmetric
2.3.3 Instance-dependent
2.4: Datasets
2.4.1 Computer Vision Datasets
2.4.1.1 Synthetic
2.4.1.2 Real
2.5 Medical Image Analysis Datasets
3: Theoretical Aspects of Noisy-label Learning
3.1: Bias Variance Decomposition
3.1.1: Regression
3.1.2: Classification
3.2: Taxonomy of Noisy Label
3.3: PAC Learning
4: Noisy-Label Learning Techniques
4.1: Loss Function
4.1.1: Label Noise Robust Loss
4.1.2: Loss Regularisation
4.1.3: Loss Reweighting
4.1.4: Loss Redesign
4.1.5: Loss Correction
4.2: Training Algorithms
4.2.1: Multi-stage Training
4.2.2: Meta-learning
4.2.3: Co-training
4.2.4: Self-training
4.2.5: Self-supervised Pre-training
4.2.6: Semi-supervised Learning
4.2.7: Expectation Maximization
4.3: Data Processing
4.3.1: Adversarial training
4.3.2: Label Cleaning
4.3.3: Sample Weighting
4.3.4: Sample Selection
4.3.5: Label Smoothing
4.4: Model Architecture
4.4.1: Adaptation Layer
4.4.2: Multiple Classifiers
4.4.3: Combining Generative and Discriminative models
5: Benchmarks, Methods, Results and Code
5.1 Closed Set Label Noise Problems
5.2 Open Set Label Noise Problems
5.3 Imbalanced Noisy-Label Problems
5.4 Imbalanced Noisy Multi-label Problems
5.5 Noisy-Label Segmentation Problems
5.6 How to Discover Future State-of-the-art Methods
6: Conclusion and Final Considerations
6.1 Conclusions
6.2 Future Work
- Edition: 1
- Published: February 23, 2024
- No. of pages (Paperback): 312
- Imprint: Academic Press
- Language: English
- Paperback ISBN: 9780443154416
- eBook ISBN: 9780443154423
GC
Gustavo Carneiro
Professor Gustavo Carneiro, Artificial Intelligence and Machine Learning, University of Surrey, UK.
Affiliations and expertise
Professor of AI and Machine Learning, Centre for Vision, Speech and Signal Processing (CVSSP), Surrey Institute for People-centred Artificial Intelligence, Department of Electrical and Electronic Engineering, The University of Surrey, UKRead Machine Learning with Noisy Labels on ScienceDirect