Learning and evaluation in presence of Non-i.i.d. label noise

Nico Görnitz, Anne K. Porbadnigk, Alexander Binder, Claudia Sannelli, Mikio Braun, Klaus Muller, Marius Kloft

Research output: Contribution to journalArticle

10 Citations (Scopus)

Abstract

In many real-world applications, the simplified assumption of independent and identically distributed noise breaks down, and labels can have structured, systematic noise. For example, in brain-computer interface applications, training data is often the result of lengthy experimental sessions, where the attention levels of participants can change over the course of the experiment. In such application cases, structured label noise will cause problems because most machine learning methods assume independent and identically distributed label noise. In this paper, we present a novel methodology for learning and evaluation in presence of systematic label noise. The core of which is a novel extension of support vector data description/one-class SVM that can incorporate latent variables. Controlled simulations on synthetic data and a real-world EEG experiment with 20 subjects from the domain of brain-computer-interfacing show that our method achieves accuracies that go beyond the state of the art.

Original languageEnglish
Pages (from-to)293-302
Number of pages10
JournalJournal of Machine Learning Research
Volume33
Publication statusPublished - 2014

Fingerprint

Identically distributed
Labels
Support Vector Data Description
Evaluation
Latent Variables
Synthetic Data
Real-world Applications
Experiment
Breakdown
Machine Learning
Data description
Brain computer interface
Methodology
Electroencephalography
Learning systems
Brain
Simulation
Experiments
Learning
Class

ASJC Scopus subject areas

  • Artificial Intelligence
  • Software
  • Control and Systems Engineering
  • Statistics and Probability

Cite this

Görnitz, N., Porbadnigk, A. K., Binder, A., Sannelli, C., Braun, M., Muller, K., & Kloft, M. (2014). Learning and evaluation in presence of Non-i.i.d. label noise. Journal of Machine Learning Research, 33, 293-302.

Learning and evaluation in presence of Non-i.i.d. label noise. / Görnitz, Nico; Porbadnigk, Anne K.; Binder, Alexander; Sannelli, Claudia; Braun, Mikio; Muller, Klaus; Kloft, Marius.

In: Journal of Machine Learning Research, Vol. 33, 2014, p. 293-302.

Research output: Contribution to journalArticle

Görnitz, N, Porbadnigk, AK, Binder, A, Sannelli, C, Braun, M, Muller, K & Kloft, M 2014, 'Learning and evaluation in presence of Non-i.i.d. label noise', Journal of Machine Learning Research, vol. 33, pp. 293-302.
Görnitz N, Porbadnigk AK, Binder A, Sannelli C, Braun M, Muller K et al. Learning and evaluation in presence of Non-i.i.d. label noise. Journal of Machine Learning Research. 2014;33:293-302.
Görnitz, Nico ; Porbadnigk, Anne K. ; Binder, Alexander ; Sannelli, Claudia ; Braun, Mikio ; Muller, Klaus ; Kloft, Marius. / Learning and evaluation in presence of Non-i.i.d. label noise. In: Journal of Machine Learning Research. 2014 ; Vol. 33. pp. 293-302.
@article{4ecb5150d1294ed48f903e99b20b2c80,
title = "Learning and evaluation in presence of Non-i.i.d. label noise",
abstract = "In many real-world applications, the simplified assumption of independent and identically distributed noise breaks down, and labels can have structured, systematic noise. For example, in brain-computer interface applications, training data is often the result of lengthy experimental sessions, where the attention levels of participants can change over the course of the experiment. In such application cases, structured label noise will cause problems because most machine learning methods assume independent and identically distributed label noise. In this paper, we present a novel methodology for learning and evaluation in presence of systematic label noise. The core of which is a novel extension of support vector data description/one-class SVM that can incorporate latent variables. Controlled simulations on synthetic data and a real-world EEG experiment with 20 subjects from the domain of brain-computer-interfacing show that our method achieves accuracies that go beyond the state of the art.",
author = "Nico G{\"o}rnitz and Porbadnigk, {Anne K.} and Alexander Binder and Claudia Sannelli and Mikio Braun and Klaus Muller and Marius Kloft",
year = "2014",
language = "English",
volume = "33",
pages = "293--302",
journal = "Journal of Machine Learning Research",
issn = "1532-4435",
publisher = "Microtome Publishing",

}

TY - JOUR

T1 - Learning and evaluation in presence of Non-i.i.d. label noise

AU - Görnitz, Nico

AU - Porbadnigk, Anne K.

AU - Binder, Alexander

AU - Sannelli, Claudia

AU - Braun, Mikio

AU - Muller, Klaus

AU - Kloft, Marius

PY - 2014

Y1 - 2014

N2 - In many real-world applications, the simplified assumption of independent and identically distributed noise breaks down, and labels can have structured, systematic noise. For example, in brain-computer interface applications, training data is often the result of lengthy experimental sessions, where the attention levels of participants can change over the course of the experiment. In such application cases, structured label noise will cause problems because most machine learning methods assume independent and identically distributed label noise. In this paper, we present a novel methodology for learning and evaluation in presence of systematic label noise. The core of which is a novel extension of support vector data description/one-class SVM that can incorporate latent variables. Controlled simulations on synthetic data and a real-world EEG experiment with 20 subjects from the domain of brain-computer-interfacing show that our method achieves accuracies that go beyond the state of the art.

AB - In many real-world applications, the simplified assumption of independent and identically distributed noise breaks down, and labels can have structured, systematic noise. For example, in brain-computer interface applications, training data is often the result of lengthy experimental sessions, where the attention levels of participants can change over the course of the experiment. In such application cases, structured label noise will cause problems because most machine learning methods assume independent and identically distributed label noise. In this paper, we present a novel methodology for learning and evaluation in presence of systematic label noise. The core of which is a novel extension of support vector data description/one-class SVM that can incorporate latent variables. Controlled simulations on synthetic data and a real-world EEG experiment with 20 subjects from the domain of brain-computer-interfacing show that our method achieves accuracies that go beyond the state of the art.

UR - http://www.scopus.com/inward/record.url?scp=84938246267&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84938246267&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84938246267

VL - 33

SP - 293

EP - 302

JO - Journal of Machine Learning Research

JF - Journal of Machine Learning Research

SN - 1532-4435

ER -