1: \begin{abstract}

2: In  machine learning research, the proximal gradient methods are popular for solving various optimization problems with  non-smooth  regularization. Inexact proximal gradient methods  are extremely important when   exactly solving  the proximal operator  is  time-consuming, or the proximal operator  does not have an analytic solution. However,  existing inexact proximal  gradient methods only consider  convex problems. The knowledge of inexact proximal gradient methods in the non-convex setting is very limited. % Moreover, for some machine learning models, there is still no proposed solver for exactly solving the  proximal operator.

3:  To address this challenge, in this paper, we first propose three  inexact proximal gradient algorithms, including the basic version and Nesterov's accelerated version. After that, we provide the theoretical analysis to the basic and Nesterov's accelerated versions.  The theoretical results show that our  inexact proximal gradient algorithms can have the same  convergence rates as the ones of exact proximal gradient algorithms  in the non-convex setting.

4:  Finally, we show the   applications of our inexact proximal gradient algorithms on  three representative non-convex learning problems. All experimental results confirm the  superiority of our new inexact proximal gradient algorithms. %The applications on robust OSCAR and link prediction show that,  our inexact proximal gradient algorithms are  significantly faster than the  exact ones. The application on robust Trace Lasso fills the vacancy that there is no proximal gradient algorithm for Trace Lasso.

5: \end{abstract}

6: