@prefix epid: .
@prefix dc: .
@prefix geo: .
@prefix bibo: .
@prefix epx: .
@prefix owl: .
@prefix void: .
@prefix event: .
@prefix skos: .
@prefix xsd: .
@prefix cc: .
@prefix ep: .
@prefix foaf: .
@prefix rdf: .
@prefix rdfs: .
@prefix eprel: .
@prefix dct: .
<>
foaf:primaryTopic ;
rdfs:comment "The repository administrator has not yet configured an RDF license."^^xsd:string .
rdfs:label "Baldwin.ps"^^xsd:string .
rdfs:label "preview.png"^^xsd:string .
rdfs:label "Baldwin.pdf"^^xsd:string .
rdfs:label "indexcodes.txt"^^xsd:string .
dc:format "text/html";
dc:title "HTML Summary of #1818 \n\nHow to shift bias: Lessons from the Baldwin effect\n\n";
foaf:primaryTopic .
dct:hasPart ;
ep:hasFile ;
eprel:hasVersion ;
eprel:hasVolatileVersion ;
eprel:haspreviewThumbnailVersion ;
rdf:type bibo:Document,
ep:Document;
rdfs:label "How to shift bias: Lessons from the Baldwin effect (Postscript)"^^xsd:string .
dct:hasPart ;
ep:hasFile ;
eprel:isVersionOf ;
eprel:isVolatileVersionOf ;
eprel:ispreviewThumbnailVersionOf ;
rdf:type ep:Document;
rdfs:label "How to shift bias: Lessons from the Baldwin effect (Image (PNG))"^^xsd:string .
dct:hasPart ;
ep:hasFile ;
eprel:hasIndexCodesVersion ;
eprel:hasVersion ,
;
eprel:hasVolatileVersion ,
;
eprel:haspreviewThumbnailVersion ;
rdf:type bibo:Document,
ep:Document;
rdfs:label "How to shift bias: Lessons from the Baldwin effect (PDF)"^^xsd:string .
dct:hasPart ;
ep:hasFile ;
eprel:isIndexCodesVersionOf ;
eprel:isVersionOf ;
eprel:isVolatileVersionOf ;
rdf:type ep:Document;
rdfs:label "How to shift bias: Lessons from the Baldwin effect (Indexer Terms)"^^xsd:string .
rdf:_1 .
bibo:abstract "An inductive learning algorithm takes a set of data as input and generates a hypothesis as\noutput. A set of data is typically consistent with an infinite number of hypotheses;\ntherefore, there must be factors other than the data that determine the output of the\nlearning algorithm. In machine learning, these other factors are called the bias of the\nlearner. Classical learning algorithms have a fixed bias, implicit in their design. Recently\ndeveloped learning algorithms dynamically adjust their bias as they search for a\nhypothesis. Algorithms that shift bias in this manner are not as well understood as\nclassical algorithms. In this paper, we show that the Baldwin effect has implications for\nthe design and analysis of bias shifting algorithms. The Baldwin effect was proposed in\n1896, to explain how phenomena that might appear to require Lamarckian evolution\n(inheritance of acquired characteristics) can arise from purely Darwinian evolution.\nHinton and Nowlan presented a computational model of the Baldwin effect in 1987. We\nexplore a variation on their model, which we constructed explicitly to illustrate the lessons\nthat the Baldwin effect has for research in bias shifting algorithms. The main lesson is that\nit appears that a good strategy for shift of bias in a learning algorithm is to begin with a\nweak bias and gradually shift to a strong bias."^^xsd:string;
bibo:authorList ;
bibo:issue "3";
bibo:status ,
;
bibo:volume "4";
dct:creator ;
dct:date "1996";
dct:isPartOf ,
;
dct:subject ,
,
;
dct:title "How to shift bias: Lessons from the Baldwin effect"^^xsd:string;
ep:hasDocument ,
,
,
;
rdf:type bibo:Article,
ep:EPrint,
ep:JournalpEPrint;
rdfs:seeAlso .
rdf:type skos:Concept;
skos:prefLabel "Evolution" .
rdf:type skos:Concept;
skos:prefLabel "Machine Learning" .
rdf:type skos:Concept;
skos:prefLabel "Statistical Models" .
foaf:familyName "Turney"^^xsd:string;
foaf:givenName "Peter D."^^xsd:string;
foaf:name "Peter D. Turney"^^xsd:string;
rdf:type foaf:Person .
foaf:name "Evolutionary Computation"^^xsd:string;
rdf:type bibo:Collection .