The Peaking Phenomenon in Semi-supervised Learning

Research output: Chapter in Book/Conference proceedings/Edited volumeConference contributionScientificpeer-review

2 Citations (Scopus)

Abstract

For the supervised least squares classifier, when the number of training objects is smaller than the dimensionality of the data, adding more data to the training set may first increase the error rate before decreasing it. This, possibly counterintuitive, phenomenon is known as peaking. In this work, we observe that a similar but more pronounced version of this phenomenon also occurs in the semi-supervised setting, where instead of labeled objects, unlabeled objects are added to the training set. We explain why the learning curve has a more steep incline and a more gradual decline in this setting through simulation studies and by applying an approximation of the learning curve based on the work by Raudys and Duin.
Original languageEnglish
Title of host publicationStructural, Syntactic, and Statistical Pattern Recognition
Subtitle of host publicationJoint IAPR International Workshop, S+SSPR 2016, proceedings
EditorsA. Robles-Kelly, Marco Loog, B. Biggio, F. Escolano, R. Wilson
Place of PublicationCham
PublisherSpringer
Pages299-309
Number of pages11
ISBN (Electronic)978-3-319-49055-7
ISBN (Print)978-3-319-49054-0
DOIs
Publication statusPublished - 2016
EventSSPR Joint IAPR International Workshops on Statistical Techniques in Pattern Recognition (SPR) and Structural and Syntactic Pattern Recognition (SSPR) - Mérida, Mexico
Duration: 29 Nov 20162 Dec 2016

Publication series

NameLecture Notes in Computer Science
Volume10029
ISSN (Print)0302-9743

Workshop

WorkshopSSPR Joint IAPR International Workshops on Statistical Techniques in Pattern Recognition (SPR) and Structural and Syntactic Pattern Recognition (SSPR)
CountryMexico
CityMérida
Period29/11/162/12/16

Keywords

  • Semi-supervised learning
  • Peaking
  • Least squares classfier
  • Pseudo-inverse

Fingerprint Dive into the research topics of 'The Peaking Phenomenon in Semi-supervised Learning'. Together they form a unique fingerprint.

Cite this