Abstract | ||
---|---|---|
This paper describes an application of the Prototype-based Min- imum Error Classification (PBMEC) to the offline recognition of handwritten digits. The PBMEC uses a set of prototypes to repre- sent each digit along with an Lν-norm of distances as the decod- ing scheme. Optimization of the system is based on the Minimum Classification Error (MCE) criterion. In this paper, we introduce a new clustering criterion adapted to the PBMEC structure that minimizes an Lν-norm-based distortion measure. The new clus- tering algorithm can generate a smaller number of prototypes than the standard k-means with no loss in accuracy. It is also shown that the PBMEC trained with the MCE can achieve over 42% im- provement from the baseline k-means process and requires only 28Kb storage to match the performance of a 1.46MB-sized k-NN classifier. |
Year | DOI | Venue |
---|---|---|
2004 | 10.1109/ICASSP.2004.1327243 | ICASSP (5) |
Keywords | Field | DocType |
training data,image classification,k means,learning artificial intelligence,application software,prototypes,optimization,minimisation,handwriting recognition,clustering algorithms | Pattern recognition,Computer science,Handwriting recognition,Minimisation (psychology),Artificial intelligence,Decoding methods,Cluster analysis,Contextual image classification,Application software,Classifier (linguistics),Distortion | Conference |
Citations | PageRank | References |
1 | 0.47 | 6 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Roongroj Nopsuwanchai | 1 | 13 | 3.12 |
Alain Biem | 2 | 288 | 18.64 |