Classification with asymmetric label noise: Consistency and maximal denoising
- Univ. of Potsdam (Germany). Inst. for Mathematics
- Pennsylvania State Univ., University Park, PA (United States). Dept. of Mechanical and Nuclear Engineering
- Univ. of Utah, Salt Lake City, UT (United States). Dept. of Mathematics
- Univ. of Michigan, Ann Arbor, MI (United States). Dept. of Nuclear Engineering and Radiological Sciences
- Univ. of Michigan, Ann Arbor, MI (United States). Dept. of Electrical and Computer Engineering, Statistics
In many real-world classification problems, the labels of training examples are randomly corrupted. Most previous theoretical work on classification with label noise assumes that the two classes are separable, that the label noise is independent of the true class label, or that the noise proportions for each class are known. In this work, we give conditions that are necessary and sufficient for the true class-conditional distributions to be identifiable. These conditions are weaker than those analyzed previously, and allow for the classes to be nonseparable and the noise levels to be asymmetric and unknown. The conditions essentially state that a majority of the observed labels are correct and that the true class-conditional distributions are “mutually irreducible,” a concept we introduce that limits the similarity of the two distributions. For any label noise problem, there is a unique pair of true class-conditional distributions satisfying the proposed conditions, and we argue that this pair corresponds in a certain sense to maximal denoising of the observed distributions. Our results are facilitated by a connection to “mixture proportion estimation,” which is the problem of estimating the maximal proportion of one distribution that is present in another. We establish a novel rate of convergence result for mixture proportion estimation, and apply this to obtain consistency of a discrimination rule based on surrogate loss minimization. Experimental results on benchmark data and a nuclear particle classification problem demonstrate the efficacy of our approach. MSC 2010 subject classifications: Primary 62H30; secondary 68T10. Keywords and phrases: Classification, label noise, mixture proportion estimation, surrogate loss, consistency.
- Research Organization:
- Univ. of Michigan, Ann Arbor, MI (United States)
- Sponsoring Organization:
- USDOE National Nuclear Security Administration (NNSA), Office of Nonproliferation and Verification Research and Development (NA-22)
- Grant/Contract Number:
- NA0002534
- OSTI ID:
- 1366694
- Journal Information:
- Electronic Journal of Statistics, Vol. 10, Issue 2; ISSN 1935-7524
- Publisher:
- Institute of Mathematical Statistics (IMS) and the Bernoulli SocietyCopyright Statement
- Country of Publication:
- United States
- Language:
- English
Web of Science
Boosting in the Presence of Outliers: Adaptive Classification With Nonconvex Loss Functions
|
journal | February 2018 |
Learning to classify from impure samples with high-dimensional data
|
journal | July 2018 |
Boosting in the Presence of Outliers: Adaptive Classification With Nonconvex Loss Functions
|
text | January 2018 |
Boosting in the Presence of Outliers: Adaptive Classification With Nonconvex Loss Functions
|
text | January 2018 |
Boosting in the Presence of Outliers: Adaptive Classification with Non-convex Loss Functions
|
text | January 2017 |
| Learning to Classify from Impure Samples with High-Dimensional Data | text | January 2018 |
Similar Records
Expectation-maximization algorithms for learning a finite mixture of univariate survival time distributions from partially specified class values
Survey of Image Denoising Methods for Medical Image Classification