dc.contributorUniversidad de Chile
dc.contributorUniversité Toulouse
dc.contributorIMCCE
dc.date.accessioned2017-05-08T21:06:42Z
dc.date.accessioned2018-06-14T00:49:40Z
dc.date.available2017-05-08T21:06:42Z
dc.date.available2018-06-14T00:49:40Z
dc.date.created2017-05-08T21:06:42Z
dc.date.issued2013
dc.identifierhttp://hdl.handle.net/10533/198264
dc.identifier14STIC-01
dc.identifierF
dc.identifier.urihttp://repositorioslatinoamericanos.uchile.cl/handle/2250/1574141
dc.description.abstractThe main goal of this project is to propose, develop, and evaluate novel strategies for selecting dynamically a subset of classifiers under the base of a deep study on the behavior of an oracle for different real environments. To this end, we will need to achieve the following specific objectives: 1) To implement strategies to initialize a set of classifiers based on diversification and accuracy aspects. 2) To implement diverse strategies aiming at appropriately combining classifiers. 3) To propose novel strategies oriented to define local region or neighborhood that will be used in the selection of sets of classifiers given an input pattern. 4) To study and evaluate the behavior of the oracle in diverse real contexts related to pattern recognition in real environments. 5) To model the behavior of the oracle aiming to generate one or more classifiers that permit an accurate decision about what classifiers must be selected for the recognition of a determined input pattern. 6) To evaluate the performance of our proposal for the dynamic selection of classifiers in real applications such as handwritten recognition, signature verification, forestal species recognition, biological cell classification, medical image, classification, and x-ray image classificationThe main goal of this project is to propose, develop, and evaluate novel strategies for selecting dynamically a subset of classifiers under the base of a deep study on the behavior of an oracle for different real environments. To this end, we will need to achieve the following specific objectives: 1) To implement strategies to initialize a set of classifiers based on diversification and accuracy aspects. 2) To implement diverse strategies aiming at appropriately combining classifiers. 3) To propose novel strategies oriented to define local region or neighborhood that will be used in the selection of sets of classifiers given an input pattern. 4) To study and evaluate the behavior of the oracle in diverse real contexts related to pattern recognition in real environments. 5) To model the behavior of the oracle aiming to generate one or more classifiers that permit an accurate decision about what classifiers must be selected for the recognition of a determined input pattern. 6) To evaluate the performance of our proposal for the dynamic selection of classifiers in real applications such as handwritten recognition, signature verification, forestal species recognition, biological cell classification, medical image, classification, and x-ray image classification
dc.relationhandle/10533/198239
dc.relationhandle/10533/198084
dc.relationhandle/10533/108039
dc.rightsinfo:eu-repo/semantics/openAcces
dc.rightsAtribución-NoComercial-SinDerivadas 3.0 Chile
dc.rightshttp://creativecommons.org/licenses/by-nc-nd/3.0/cl/
dc.titleDynamic Selection of Classifiers with Application in Real Environments
dc.typeProyecto


Este ítem pertenece a la siguiente institución