÷ƒ’À;è TeX output 1995.01.18:1649‹ÿÿÿÿ ‹ú- ıW‘ì Šn ı‘ì’ó'ò"V ó3 cmbx10ÒT‘şó\ec•¦thnical›2Univ“ersit“y˜Graz˜(TUG)¤ ™š’ Ê5Departmen¦tt–2of“Medical“Informatics¡’áò¼and–2Inst.‘˜¼of“Theoretical“Computer“Science¡’³D©Graz¡’¤ã'AustriaŸ'34‘sçResearc¦thers¡‘sçó"Kñ`y ó3 cmr10ÍW‘ÿeolfgang–¦fMaass,“P²!eter“Auer,“Doris“Flotzinger.©34‘sçÒILP‘2Areas¡‘sçÍComputational–¦flearning“theory‘ÿe,“Inductivš²!e“learning,“Neural“net˜w˜orks.¦‘sçÒDescription–2of“Researc¦th¡‘sçÍThe–µgroup“concen²!trates“primarily“on“the“application“of“computational“learning“theory“to“the¡‘sçdev•²!elopmen“t–Ëâof“new“learning“algorithms,‘÷–dev•²!elopmen“t–Ëâof“hš²!ybrid“(sym˜bMŞolic“and“subsym˜bMŞolic)¡‘sçlearning–¦fsystems“and“inductiv²!e“logic“programming.¡‘-sçAmong–dƒma‘›»jor“ac•²!hiev“emen“ts,‘” the›dƒcon“tribution˜to˜the˜researc“h˜of˜the˜complexit“y˜of˜in-¡‘sçductivš²!e–;ølearning“and“to“the“theoretical“foundations“of“neural“net˜w˜orks“should“bMŞe“men˜tioned¡‘sç[2‘yš,‘ò3– k´,‘ò6“,‘ò8“]–òas“wš²!ell“as“the“solution“of“the“computational“complexit˜y“of“function“learning¡‘sç[7‘yš].‘ã—In–Rùthe“ eld“of“inductivš²!e“learning,‘~the“group“dev˜elopMŞed“and“co-dev˜elopMŞed“the“systems¡‘sçF–şÈˆA“V²!ORIT‘gÔ[4›yš]–gäand“FLORA1-3“[1˜,“5‘á~]“capable“of“learning“under“signi can²!t“concept“drift.‘ÉRe-¡‘sçcenš²!tly‘ÿe,‘[the–»÷system“TBNN‘»¯initializing“neural“nets“with“an“inductiv˜e-learning“methoMŞd“has¡‘sçb•MŞeen›¦fdev²!elop“ed˜and˜successfully˜tested˜in˜a˜medical˜domain˜[9‘yš].¡‘-sçRecen•²!t›å¶researc“h˜in“terests˜include˜the˜study˜of˜the˜computational˜complexit“y˜of˜ILP-¡‘sçorienš²!ted–¦flearning“algorithms“and“the“use“of“subsym˜bMŞolic“approac˜hes“for“theory“revision.¦‘sçÒRecen¦tt‘2Publications¡‘!íÍ[1]‘2ö M.–W=Kubat.“Conceptual“inductivš²!e“learning|The“case“of“unreliable“teac˜hers.“ó%ı': ó3 cmti10ĞA¸\rti cial¡‘2ö Intel‘Fligenc‘ÿp¹e–¦fÍ52,“169{182,“1991.¡‘!í[2]‘2ö Z.–ÈChen,‘ÊĞW.“Maass.“On-line“learning“of“rectangles.“ĞPr–ÿp¹o“c.–·5th“Aš¸\nnual“A˜CM‘°Workshop¡‘2ö on–êêComputational“L–ÿp¹e“arning‘êêThe“oryÍ,–¦f16{28,“1992.¡‘!í[3]‘2ö Z.–»TChen,‘ÀW.“Maass.“A‘»Nsolution“to“the“credit“assignmnet“problem“in“the“case“of“learn-¡‘2ö ing–™Œrectangles.“ĞPr–ÿp¹o“c.›Êœ3r“d˜Internat.˜Workshop˜on˜A¸\nalo“gic“al˜and˜Inductive˜Infer“enc“eÍ,¡‘2ö Lecture–¦fNotes“in“Arti cial“In²!telligence,“V›ÿeol.“642,“26{34,“Springer-V˜erlag,“1992.¡‘!í[4]‘2ö M.–ØKubat,‘o3I.“Krizak•²!o“v‘ÿdDa.–ØF‘ÿeorgetting“and“ageing“of“kno²!wledge“in“concept“formation.¡‘2ö ĞApplie›ÿp¹d–êêA¸\rti cial“Intel‘Fligenc˜e–¦fÍ6,“193{204,“1992.¡‘!í[5]‘2ö G.–¯ÓWidmer,‘ò.M.“Kubat.“E ectivš²!e“learning“in“dynamic“en˜vironmen˜ts“b˜y“explicit“con-¡‘2ö text–å trac²!king.“ĞPr–ÿp¹o“c.›$Eur“op“e“an˜Confer“enc“e˜on˜Machine˜L“e“arningÍ,–ôµECML‘äü'93,“227{243,¡‘2ö Vienna.–¦fSpringer-V‘ÿeerlag,“1993.¡‘!í[6]‘2ö W.–C¼Maass,‘WxG.“T‘ÿeuran.“Lo•²!w“er–C¼bšMŞound“metho˜ds“and“separation“results“for“on-line“learning¡‘2ö mošMŞdels.–¦fIn²!vited“pap˜er“for“a“sp˜ecial“issue“of“ĞMachine‘êêL–ÿp¹e“arningÍ,–¦f9,“107{145,“1992.¡‘!í[7]‘2ö P–ÿe.›ªAuer,‘ªıP“.M.˜Long,–ªıW.˜Maass,“G.J.˜W‘ÿeoMŞeginger.˜On˜the˜complexit²!y˜of˜function˜learn-¡‘2ö ing.‘æuĞPr–ÿp¹o“c.–%Ù5th“Aš¸\nnual“A˜CM‘%ÊConfer–ÿp¹enc“e–%Ùon“Computational“L–ÿp¹e“arning‘%ÙThe“oryÍ,‘öy1993‘æu(to¡‘2ö appMŞear).¡‘!í[8]‘2ö Z.–†ÆChen,‘¾İW.“Maass.“On-line“learning“of“rectangles“and“unions“of“rectangles.“In²!vited¡‘2ö papšMŞer–¦ffor“a“sp˜ecial“issue“of“ĞMachine‘êêL–ÿp¹e“arning–¦fÍ(to“app˜ear).Œ‹* ‹ú- ıW‘ì Šn ı‘ì‘!íÍ[9]‘2ö I.›÷°Iv–ÿdDano²!v“a,– M.˜Kubat,“G.˜Pfurtsc²!heller.˜A‘÷›decision-tree˜based˜neural˜net:‘€rDescription¤ ™š‘2ö and–[·a“case“study‘ÿe.“ĞPr–ÿp¹o“c.–}>11th“International“Confer–ÿp¹enc“e–}>on“Machine“L–ÿp¹e“arningÍ,‘É 1994¡‘2ö (submitted).Ÿ34‘sçÒCon•¦ttact‘2P“erson¡‘sçÍProf.‘İİW‘ÿeolfgang‘¦fMaasst¡‘sçInstitute–¦fof“theoretical“computer“science¡‘sçT‘ÿeec•²!hnical‘¦fUniv“ersit“y¡‘sçKlosterwiesgasse–¦f32“/“IMŞIa¡‘sçA-8010‘¦fGraz¡‘sçAustria¡‘sçE-mail:‘İİmaass@igi.tu-graz.ac.atŒø­ƒ’À;è‹ú-Οó'ò"V ó3 cmbx10ó%ı': ó3 cmti10ó"Kñ`y ó3 cmr10ùßßßßßßß