Network Game and Boosting

We propose an ensemble learning method called Network Boosting which combines weak learners together based on a random graph (network). A theoretic analysis based on the game theory shows that the algorithm can learn the target hypothesis asymptotically. The comparison results using several datasets of the UCI machine learning repository and synthetic data are promising and show that Network Boosting has much resistance to the noisy data than AdaBoost through the cooperation of classifiers in the classifier network.