A study of the effect of different types of noise on the precision of supervised learning techniques

By David F. Nettleton, Albert Orriols-Puig, Albert Fornells. Artificial Intelligence Review, 2010

436 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/91f19b43cedec7ba5c3796a6bc16018455c7a531

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Combating Label Noise in Deep Learning Using Abstention (2019, background): “Classification performance degrades in the presence of noise (Nettleton et al., 2010), with label noise shown to be more harmful than feature noise (Zhu & Wu, 2004).”
CleanNet: Transfer Learning for Scalable Image Classifier Training with Label Noise (2017, background): “Yet many studies have shown that label noise can affect accuracy of the induced classifiers significantly [7, 19, 22, 27], making it desirable to develop algorithms for learning in presence of label noise.”
Robot Learning in Homes: Improving Generalization and Reducing Dataset Bias (2018, background): “[36] show that training models from noisy data detrimentally impacts performance.”
Supervised machine learning: A brief primer (2020, background): “Label noise has been shown to lead to inaccurate predictor selection and rankings (Frénay et al., 2014; Gerlach & Stamey, 2007; Shanab et al., 2012) and to decrease prediction performance (Lachenbruch, 1966; Nettleton et al., 2010; Wilson & Martinez, 2000).”
Using Self-Supervised Learning Can Improve Model Robustness and Uncertainty (2019, background): “Another robustness problem is learning in the presence of corrupted labels [Nettleton et al., 2010, Patrini et al., 2017].”
Effects of Distance Measure Choice on K-Nearest Neighbor Classifier Performance: A Review (2017, neutral): “38 Noisy data are a corrupted form of data in some way, which leads to partial alteration of the data values.”
Training Convolutional Networks with Noisy Labels (2014, background): “In any classification model, a degradation in performance is inevitable when there is noise in the training data (Nettleton et al., 2010; Pechenizkiy et al., 2006).”
Iterative Learning with Open-set Noisy Labels (2018, background): “The presence of noisy labels for training samples may adversely affect representation learning and deteriorate prediction performance [27].”
Using Trusted Data to Train Deep Networks on Labels Corrupted by Severe Noise (2018, background): “The performance of machine learning systems reliant on labeled data has been shown to degrade noticeably in the presence of label noise [15, 17].”
Seeing through the Human Reporting Bias: Visual Classifiers from Noisy Human-Centric Labels (2015, background): “Our goal is to train visually grounded image classifiers for a set of visual concepts w ∈ W (e.g., banana, yellow, zebra) using images and their human-centric annotations.”
Deep Self-Learning From Noisy Labels (2019, background): “But collecting large-scale datasets with precise annotations is expensive and time-consuming, preventing DNNs from being employed in real-world noisy scenarios.”
Image Classification with Deep Learning in the Presence of Noisy Labels: A Survey (2019, background): “tr (Ilkay Ulusoy) 1ASELSAN, Ankara 2Middle East Technical University, Electrical-Electronics Engineering, Ankara be more harmful [14, 17] and shown to deteriorate the performance of classification systems in a broad range of problems [14, 18, 19].”
Systematic Poisoning Attacks on and Defenses for Machine Learning in Healthcare (2015, background): “We note that the differences in robustness or vulnerability of the machine-learning algorithms to our proposed attacks are analogous to the dissimilar effects of noise on different machinelearning algorithms [10], [11].”
Disease Diagnosis in Smart Healthcare: Innovation, Technologies and Applications (2017, background): “Some of the difficulties are: large dimensionality of feature vectors [50], bias/variance dilemma [51], input and output noise [52], large-scale training data [53], data heterogeneity [54], data redundancy [55] and non-linearity among features [56].”
Learning from Noisy Labels with Deep Neural Networks (2014, methodology): “In any classiﬁcation model, degradation of performance is inevitable when there is noise in training labels [13, 15].”
Learning from massive noisy labeled data for image classification (2015, background): “noisy labels could adversely impact the classification accuracy of the induced classifiers [20, 22, 31].”
ModelTracker: Redesigning Performance Analysis Tools for Machine Learning (2015, background): “Mislabeled data (often referred to as label or class noise) frequently occurs during data collection in machine learning [15] and can be resolved by locating and correcting the incorrect labels.”
Training deep neural-networks based on unreliable labels (2016, background): “The presence of class label noise inherent to training samples has been reported to deteriorate the performance of even the best classifiers in a broad range of classification problems [1][2].”