Next: Injecting Randomness
Up: Ensembles
Previous: Manipulating the Input Features
- Error Correcting Output Coding - randomly partition K classes
into two subsets A and B, learn a classifier, repeat the process L
times
- each member of each class receives a vote and the class with the
most votes is the prediction of the ensemble
- methods for designing good error-correcting codes can be applied
- has been combined with Adaboost
- ECOC has also been combined with feature-selection
Patricia Jean Riddle
Wed Jun 23 13:06:34 NZST 1999