1: \begin{abstract}

2: We consider concave minimization problems over nonconvex sets.

3: Optimization problems with this structure arise in sparse

4: principal component analysis.

5: We analyze both a gradient projection algorithm and an approximate

6: Newton algorithm where the Hessian approximation is a multiple of the identity.

7: Convergence results are established.

8: In numerical experiments arising in sparse principal component analysis,

9: it is seen that the performance of the

10: gradient projection algorithm is very similar to that of the truncated

11: power method and the generalized power method.

12: In some cases, the approximate Newton algorithm with

13: a Barzilai-Borwein (BB) Hessian approximation

14: can be substantially faster than the other algorithms, and can converge

15: to a better solution.

16: \end{abstract}

17: