Neural learning using AdaBoost

This paper describes a committee of neural networks submitted to the IJCNN 2001 generalization ability challenge (GAC) competition and a number of implementation issues with a focus on the generalization problem. The committee of neural networks was generated using the well-known AdaBoost, which is a general method for improving the performance of any learning algorithm that consistently generates classifiers that perform better than random guessing. We also discuss the feature selection and various experiments conducted in the hope of finding a neural network architecture that can generalize correctly in the blind test data used in the GAC competition.