1: \begin{abstract}

2: 	A broad class of optimization problems can be cast in composite form,

3: 	that is,

4: 	considering the minimization

5: 	of the composition of a lower semicontinuous function

6: 	with a differentiable mapping.

7: 	This paper investigates the versatile template of composite optimization 

8: 	without any convexity assumptions.

9: 	First- and second-order optimality conditions are discussed.

10: 	We highlight the difficulties that stem from the lack of convexity

11: 	when dealing with necessary conditions in a Lagrangian framework

12: 	and when considering error bounds.

13: 	Building upon these characterizations,

14: 	a local convergence analysis is delineated for a recently developed augmented Lagrangian method,

15: 	deriving rates of convergence in the fully nonconvex setting.

16: 	

17: 	\begin{keywords}

18: 		\TheKeywords

19: 	\end{keywords}

20: 	

21: 	\begin{subclass}

22: 		\TheAMSsubj

23: 	\end{subclass}

24: \end{abstract}

25: