next up previous
Next: Conclusions Up: No Title Previous: Support Vector Machines

Numerical experiments

Applying the above algorithms to some standard polychotomic databases, the following results have been obtained:


tabular360

The first column labeled C4.5 is the result of the decision tree algorithm C4.5 applied to the original data from tex2html_wrap_inline695 and is used as reference.

Each number in this table is an average over 60 runs. In each run, the total dataset has been split into 3 equal parts. The first one is used to train the classifiers learning each dichotomy tex2html_wrap_inline697 (n.b. the decision tree algorithm C4.5 was used for these tasks); the second is used to determine the mapping m, while the third is used for testing the overall system.



Eddy Mayoraz
Tue Mar 31 18:35:24 MET DST 1998