Constraint awareness in balanced ensemble learning

By weakening the error signals on the learned data points and enforcing the error signals on those not-yet-learned data points, balanced ensemble learning was developed from negative correlation learning. Although balanced ensemble learning could learn faster and better than negative correlation learning, it also carried higher risk of overfitting in case of having limited number of training data points. If there could be enough data points, such risk could be removed away. In this paper, balanced ensemble learning with constraint was developed through bringing random data points in training. Experimental results were carried out to analyze how such constraint awareness could guide the learning trace, and limit the variances in balanced ensemble learning.