Analogical classification: Handling numerical data

Publication Type:
Conference Proceeding
Citation:
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2014, 8720 pp. 66 - 79
Issue Date:
2014-01-01
Metrics:
Full metadata record
Files in This Item:
Filename Description Size
10.1007_978-3-319-11508-5.pdfPublished version4.38 MB
Adobe PDF
© Springer International Publishing Switzerland 2014. The formal modeling of analogical proportions (i.e., statements of the form “a is to b as c is to d”) has led in the last past years to different proposals for classification algorithms, which have been quite successful on benchmarks where data are described by binary or nominal features. As far as we know and up to one exception, numerical data have never been considered. We propose here a new algorithm for handling numerical data. Starting from multiple-valued logical modelings of analogical proportions, more or less strongly encoding the idea that the change from a to b is the same as the change from c to d, we investigate different implementations leading to very good results on classical benchmarks.
Please use this identifier to cite or link to this item: