Title :
A data-dependent distance measure for transductive instance-based learning
Author :
Lundell, Jared ; Ventura, Dan
Author_Institution :
Brigham Young Univ., Provo
Abstract :
We consider learning in a transductive setting using instance-based learning (k-NN) and present a method for constructing a data-dependent distance "metric" using both labeled training data as well as available unlabeled data (that is to be classified by the model). This new data-driven measure of distance is empirically studied in the context of various instance-based models and is shown to reduce error (compared to traditional models) under certain learning conditions. Generalizations and improvements are suggested.
Keywords :
learning (artificial intelligence); matrix algebra; pattern clustering; data clustering; data-dependent distance measure matrix; labeled training data; transductive instance-based learning; Computer errors; Computer science; Context modeling; Data acquisition; Labeling; Machine learning; Semisupervised learning; Support vector machines; Training data; Upper bound;
Conference_Titel :
Systems, Man and Cybernetics, 2007. ISIC. IEEE International Conference on
Conference_Location :
Montreal, Que.
Print_ISBN :
978-1-4244-0990-7
Electronic_ISBN :
978-1-4244-0991-4
DOI :
10.1109/ICSMC.2007.4414133
