The research done on this work majorly focuses on one question,
“What is it then that distinguishes neural networks that generalize well from those that don’t?”.
Generalization, Model Capacity & Regularization are some of the important techniques used for this research work. Generalization error is defined as the difference between training error & the testing error. As the model gets trained on more and more datasets, the difference between training error & testing error keeps on increasing. This leads to the problem of over-fitting.