1: \begin{abstract}

2: In this note, we consider the line search for a class of abstract nonconvex algorithm which have been deeply studied in the Kurdyka-{\L}ojasiewicz theory. We provide a weak convergence result of the line search in general. When the objective function satisfies the Kurdyka-{\L}ojasiewicz property and some certain assumption, a global convergence result can be derived. An application  is presented for the $\ell_0$ regularized least square minimization in the end of the paper.

3: \end{abstract}