590 research outputs found
UJM at CLEF in Author Verification based on optimized classification trees
No full texthttp://ceur-ws.org/Vol-1180/CLEF2014wn-Pan-FreryEt2014.pdfInternational audienceThis article describes our proposal for the Author Identification task in the PAN CLEF Challenge 2014. We have adopted a machine learning ap- proach based on several representations of the texts and on optimized decision trees which have as entry various attributes and which are learned for every train- ing corpus separately for this classification task. Our method ranked us at the 2nd place with an overall AUC of 70.7%, and C@1 of 68.4% and, between the 1st and the 6th place on the six corpora
UJM at CLEF in Author Verification based on optimized classification trees
No full texthttp://ceur-ws.org/Vol-1180/CLEF2014wn-Pan-FreryEt2014.pdfInternational audienceThis article describes our proposal for the Author Identification task in the PAN CLEF Challenge 2014. We have adopted a machine learning ap- proach based on several representations of the texts and on optimized decision trees which have as entry various attributes and which are learned for every train- ing corpus separately for this classification task. Our method ranked us at the 2nd place with an overall AUC of 70.7%, and C@1 of 68.4% and, between the 1st and the 6th place on the six corpora
UJM at CLEF in Author Verification based on optimized classification trees. Notebook for PAN at CLEF 2014
No full textUJM at CLEF in Author Verification based on optimized classification trees. Notebook for PAN at CLEF 2014
No full textRepartition dynamique de donnees regulieres pour des machines MIMD homogenes a memoire distribuee
No full texthttp://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.39.6074&rep=rep1&type=pdfInternational audienceThis paper reports a simple strategy for dynamically load balancing of regular data on homogeneous Distributed Memory MIMD computers with static interconnexion network. That is data in which each unit can be processed independently and requires the same time on each processor. This strategy is based upon a global knowledge of the initial distribution of the data' load obtained by a generalized prefix computation on a spanning tree of the graph of the interconnexion network of the processors; this strategy is then usable for any topology. We prove also that the data exchangesinduced by this strategy lead to the balance of the data' load in time O(min P kP kK(P )), where P is a maximum weight path of the spanning tree, kP k its length and K(P ) the maximum weight of its arcs
A New Early Ripening Table Grape Cultivar Obtained in Pietroasa Research Station: Vitis vinifera L. cv. Mihaela
No full textAbstractThe evolution of the internal Romanian market for table grapes in recent years suggests that there are occasional problems with the supply, sometimes requiring imports, as well as a need for early maturing varieties, which can bring important benefits for both producers and consumers. This paper presents some data regarding a promising new variety of table grapes, Mihaela, obtained at the Pietroasa Research Station located in Buzău county, Romania, by crossing of Vitis vinifera varieties Cardinal and Coarnă neagră, the last being the maternal ancestor. Presented data includes ampelographic features, agrobiological and technological traits for the new Mihaela cultivar, as well as for its parent varieties. The results recorded thus far present Mihaela as a valuable early ripening variety which could be a good addition to the table grapes assortment of many producers
A Complex Mining Process about Air Quality
No full texthttp://www.atlantis-press.com/php/download_paper.php?id=9649International audienceIn this paper we present a mining project about extracting knowledge from public documents concerning air pollution. Our collection contains annual reports about air quality, acid rains, climatological conditions in the large area of Mexico City. These reports contain reliable data and are generated by the Department of Environment, they are in a printable format (.pdf file) with number of pages, table of content, textual information, numerical information in tables, images. For a human being it is impossible to read the whole collection during a relatively short period (a few days or weeks) and understand the content of them. An automatic box of tools able to extract knowledge, to quick retrieve important term, to answer some exact questions about precise climate parameters would be an important help for lecturers. We will describe our project based upon a text and data mining process; the aims of the complex process are extract frequent temporal pattern, to extract association rules, to integrate also some information retrieval simple tools. In parallel, some data mining techniques will be used to detect the same types of data presented in every report and then to extract a numerical datamart containing climatological data structured by month, year, geographical area. The datamart will be analyzed also. The main steps of our mining process are: preparing documents (cleaning, removing images, table of contents, footnotes), transforming in structured document (in a XML format with a precise DTD), indexing, various algorithms and methods of mining, visualising results and validating knowledge. We think also that our methodology will concern also other collections of the same category : reliable data and informations presented in huge periodical reports
- …
