Our goal is to fit the multiple instances (or structures) of a generic model existing in data. Here we propose a novel model selection scheme to estimate the number of genuine str...
—This paper describes effective object function design for combining on-line and off-line character recognizers for on-line handwritten Japanese text recognition. We combine on-l...
Many semi-supervised learning algorithms only
deal with binary classification. Their extension to the
multi-class problem is usually obtained by repeatedly
solving a set of bina...
The segmentation of anatomical structures has been traditionally formulated as a perceptual grouping task, and solved through clustering and variational approaches. However, such ...
Bogdan Georgescu, Xiang Sean Zhou, Dorin Comaniciu...
Abstract. This paper describes an efficient method to construct reliable machine learning applications in peer-to-peer (P2P) networks by building ensemble based meta methods. We co...