β Input |
ββA training set ; A base learner ; Number of iterations ; A new data point |
ββto be classified. |
β Training Phase |
ββInitialization: Set the weight distribution over as . |
ββFor |
ββ(1) According to the distribution , draw training instances at random from with |
βββreplacement to compose a new set . |
ββ(2) Provide as the input of to train a classifier , and then compute the weighted |
βββtraining error of as β β β β β β |
βββββββββ, ββββ(1) |
βββwhere takes value 1 or 0 depending on whether the th training instance is |
βββmisclassified or by or not. |
ββ(3) If or , then set and abort loop. |
ββ(4) Let ). |
ββ(5) Update the weight distribution over as |
ββββββββββββββ(2) |
βββwhere is a normalization factor being chosen so that is a probability |
βββdistribution over . |
ββEndfor |
β Output |
βββ The class label for predicted by the ensemble classifier as |
βββββββββββββ. |