1: \begin{abstract}

2: %% Text of abstract

3: % In this paper, we propose an effective method to improve the model generalization by the coupled tensor norm regularization. The novel regularization could enable the model output feature coupled with the data input to lie in a low-dimensional manifold, which helps us to reduce overfitting. We show that this regularization term is convex, differentiable, and gradient Lipschitz continuous for logistic regression, while nonconvex and nonsmooth for deep neural networks. %We demonstrate that the coupled tensor norm regularization could enforce the model output feature to lie in a low-dimensional manifold coupled with the data input, which helps us to reduce overfitting. 

4: % %Our proposed regularization is data-dependent and efficient especially for small data problems. 

5: % As a data-dependent regularization, this approach is efficient especially for small data problems. We leverage the first-order method to fit in the gradient descent framework and analyze their global 

6: % convergence. The results of numerical experiments on nine datasets demonstrate that our proposed method could improve the generalization performance of both multinomial logistic regression and deep neural network models. %on nine datasets.

7: 

8: In this paper, we propose a coupled tensor norm regularization that could enable the model output feature and the data input to lie in a low-dimensional manifold, which helps us to reduce overfitting. We show this regularization term is convex, differentiable, and gradient Lipschitz continuous for logistic regression, while nonconvex and nonsmooth for deep neural networks. We further analyze the convergence of the first-order method for solving this model. The numerical experiments demonstrate that our method is efficient.

9: 

10: \end{abstract}

11: