Title :
Research on the characteristic of the probabilistic outputs via random ECOC
Author :
Wang, Xiaodan ; Zhou, Jindeng
Author_Institution :
Dept. of Comput. Eng., Air Force Eng. Univ., Xi´´an, China
Abstract :
Error-correcting output codes (ECOC) is an efficient method to solve the problem of multi-class classification, and how to get the accurate probability estimation via ECOC is also attractive for many applications. The characteristic of the probabilistic outputs via random ECOC is analyzed in this paper, and the proof is given for that the probability output based on random ECOC is the unbiased estimation. Experimental results validate the correctness of our conclusion.
Keywords :
error correction codes; estimation theory; probability; error-correcting output codes; multiclass classification problem; probabilistic output characteristics; probability estimation; random ECOC; Biological system modeling; Encoding; Estimation; Learning; Noise; Probabilistic logic; Support vector machines; Error-correcting output codes; Multi-class classification; Probability estimation;
Conference_Titel :
Progress in Informatics and Computing (PIC), 2010 IEEE International Conference on
Conference_Location :
Shanghai
Print_ISBN :
978-1-4244-6788-4
DOI :
10.1109/PIC.2010.5687435
