<?xml version="1.0" encoding="utf-8"?>
<?xml-stylesheet type="text/xsl" href="style/detail_T.xsl"?>
<bibitem type="C">   <ARLID>0337806</ARLID> <utime>20240103192946.3</utime><mtime>20100128235959.9</mtime>         <title language="eng" primary="1">On the Over-Fitting Problem of Complex Feature Selection Methods</title>  <specification> <page_count>6 s.</page_count> <media_type>CD-ROM</media_type> </specification>   <serial><ARLID>cav_un_epca*0337805</ARLID><title>Proc. 5th International Computer Engineering Conference - A better Information Society Through the e@</title><part_num/><part_title>Machine Intelligence and Web Applications</part_title><page_num>12-17</page_num><publisher><place>Káhira</place><name>Cairo University</name><year>2009</year></publisher></serial>   <title language="cze" primary="0">O problému přetrénování komplexních metod výběru příznaků</title>    <keyword>feature selection</keyword>   <keyword>overfitting</keyword>   <keyword>overselection</keyword>    <author primary="1"> <ARLID>cav_un_auth*0101197</ARLID> <name1>Somol</name1> <name2>Petr</name2> <full_dept language="cz">Rozpoznávání obrazu</full_dept> <full_dept language="eng">Department of Pattern Recognition</full_dept> <department language="cz">RO</department> <department language="eng">RO</department> <institution>UTIA-B</institution> <full_dept>Department of Pattern Recognition</full_dept>  <fullinstit>Ústav teorie informace a automatizace AV ČR, v. v. i.</fullinstit> </author> <author primary="0"> <ARLID>cav_un_auth*0101171</ARLID> <name1>Novovičová</name1> <name2>Jana</name2> <full_dept language="cz">Rozpoznávání obrazu</full_dept> <full_dept>Department of Pattern Recognition</full_dept> <department language="cz">RO</department> <department>RO</department> <institution>UTIA-B</institution>  <fullinstit>Ústav teorie informace a automatizace AV ČR, v. v. i.</fullinstit> </author> <author primary="0"> <ARLID>cav_un_auth*0101182</ARLID> <name1>Pudil</name1> <name2>Pavel</name2> <full_dept language="cz">Rozpoznávání obrazu</full_dept> <full_dept>Department of Pattern Recognition</full_dept> <department language="cz">RO</department> <department>RO</department> <institution>UTIA-B</institution> <full_dept>Department of Pattern Recognition</full_dept>  <fullinstit>Ústav teorie informace a automatizace AV ČR, v. v. i.</fullinstit> </author>   <source> <url>http://library.utia.cas.cz/separaty/2010/RO/somol-on the over-fitting problem of complex feature selection methods.pdf</url> </source>        <cas_special> <project> <project_id>2C06019</project_id> <agency>GA MŠk</agency> <country>CZ</country> <ARLID>cav_un_auth*0216518</ARLID> </project> <project> <project_id>1M0572</project_id> <agency>GA MŠk</agency> <ARLID>cav_un_auth*0001814</ARLID> </project> <project> <project_id>GA102/07/1594</project_id> <agency>GA ČR</agency> <ARLID>cav_un_auth*0228611</ARLID> </project> <project> <project_id>GA102/08/0593</project_id> <agency>GA ČR</agency> <ARLID>cav_un_auth*0239567</ARLID> </project> <research> <research_id>CEZ:AV0Z10750506</research_id> </research>  <abstract language="eng" primary="1">One of the hot topics discussed recently in relation to machine learning is the question of actual performance of modern feature selection methods. Feature selection has been a highly active area of research in recent years due to its potential to improve both the performance and economy of automatic decision systems  in various applicational fields, including medicine, image analysis,  remote sensing, economics etc. The number of available methods and methodologies has grown rapidly throughout recent years while promising important improvements. Yet recently many authors put this development in question, claiming that simpler older tools are actually better than complex modern ones – which, despite promises, are claimed to actually fail in real-world applications. We investigate this question, show several illustrative  examples and draw several conclusions and recommendations regarding feature selection methods’ expectable performance.</abstract> <abstract language="cze" primary="0">Jedno z aktuálních témat ve strojovém učení je otázka skutečné výkonnosti moderních metod výběru příznaků. Výběr příznaků byl v posledních letech v centru pozornosti pro svou schopnost zlepšovat nejen úspěšnost ale i snižovat náročnost systémů automatického rozhodování v různých oblastech - v medicíně, zpracování obrazových dat, ekonomii atd. Počet dostupných metod a matodologií stále narůstá přičemž nové metody slibují další zlepšení. Tento vývoj byl nicméně nedávno zpochybněn řadou autorů s poukazem na pozorování, že starší metody se ve skutečnosti chovají lépe v reálných aplikacích. V této práci zkoumáme tuto otázku a na několika ilustrativních příkladech podáváme doporučení ohledně očekávatelného výkonu různých metod.</abstract>  <action target="WRD"> <ARLID>cav_un_auth*0259324</ARLID> <name>5th International Computer Engineering Conference - A better Information Society Through the e@</name> <place>Káhira</place> <dates>27.12.2009-28.12.2009</dates>  <country>EG</country> </action>    <reportyear>2010</reportyear>  <RIV>BD</RIV>      <permalink>http://hdl.handle.net/11104/0181723</permalink>        <arlyear>2009</arlyear>       <unknown tag="mrcbU63"> cav_un_epca*0337805 Proc. 5th International Computer Engineering Conference - A better Information Society Through the e@ Machine Intelligence and Web Applications 12 17 Káhira Cairo University 2009 </unknown> </cas_special> </bibitem>