Least squares approach to K-SVCR multi-class classification with its applications

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F22%3A10453335" target="_blank" >RIV/00216208:11320/22:10453335 - isvavai.cz</a>

Výsledek na webu

<a href="https://verso.is.cuni.cz/pub/verso.fpl?fname=obd_publikace_handle&handle=jpsldcMcJG" target="_blank" >https://verso.is.cuni.cz/pub/verso.fpl?fname=obd_publikace_handle&handle=jpsldcMcJG</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/s10472-021-09747-1" target="_blank" >10.1007/s10472-021-09747-1</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Least squares approach to K-SVCR multi-class classification with its applications

Popis výsledku v původním jazyce

The support vector classification-regression machine for K-class classification (K-SVCR) is a novel multi-class classification method based on the &quot;1-versus-1-versus-rest&quot; structure. In this paper, we propose a least squares version of K-SVCR named LSK-SVCR. Similarly to the K-SVCR algorithm, this method assesses all the training data into a &quot;1-versus-1-versus-rest&quot; structure, so that the algorithm generates ternary outputs {- 1,0,+ 1}. In LSK-SVCR, the solution of the primal problem is computed by solving only one system of linear equations instead of solving the dual problem, which is a convex quadratic programming problem in K-SVCR. Experimental results on several benchmark, MC-NDC, and handwritten digit recognition data sets show that not only does the LSK-SVCR have better performance in the aspects of classification accuracy to that of K-SVCR and Twin-KSVC algorithms but also has remarkably higher learning speed.

Název v anglickém jazyce

Least squares approach to K-SVCR multi-class classification with its applications

Popis výsledku anglicky

The support vector classification-regression machine for K-class classification (K-SVCR) is a novel multi-class classification method based on the &quot;1-versus-1-versus-rest&quot; structure. In this paper, we propose a least squares version of K-SVCR named LSK-SVCR. Similarly to the K-SVCR algorithm, this method assesses all the training data into a &quot;1-versus-1-versus-rest&quot; structure, so that the algorithm generates ternary outputs {- 1,0,+ 1}. In LSK-SVCR, the solution of the primal problem is computed by solving only one system of linear equations instead of solving the dual problem, which is a convex quadratic programming problem in K-SVCR. Experimental results on several benchmark, MC-NDC, and handwritten digit recognition data sets show that not only does the LSK-SVCR have better performance in the aspects of classification accuracy to that of K-SVCR and Twin-KSVC algorithms but also has remarkably higher learning speed.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

50201 - Economic Theory

Projekt

<a href="/cs/project/GA18-04735S" target="_blank" >GA18-04735S: Nové přístupy pro relaxační a aproximační techniky v deterministické globální optimalizaci</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2022

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Annals of Mathematics and Artificial Intelligence

ISSN

1012-2443

e-ISSN

1573-7470

Svazek periodika

90

Číslo periodika v rámci svazku

7-9

Stát vydavatele periodika

NL - Nizozemsko

Počet stran výsledku

20

Strana od-do

873-892

Kód UT WoS článku

000663987900001

EID výsledku v databázi Scopus

2-s2.0-85123104517