Reducing Overfitting Problem in Machine Learning Using Novel L1/4 Regularization Method

Abstract

The Machine learning model has two problems, they are Overfitting and Under-fitting. Underfitting is a statistical model or a machine learning algorithm, it cannot capture the underlying trend of the data. A statistical model is said to be overfitted when it has been trained with more data. When the model is trained on fewer features, the machine will be too biased, and then the model gets under fitting problems. So, it has been required to train the model on more features and there is one more problem that occurs. To reduce the overfitting problem, regularization functions and data augmentation are used. Lasso shrinks the less important feature's coefficient to zero thus removing some feature altogether. L2 regularization, on the other hand, does not remove most of the features. A novel regularization method is proposed to overcome these problems.