1: \begin{abstract}

2: Proximal distance algorithms combine the classical penalty method of

3: constrained minimization with distance majorization. If $f(\bx)$ is

4: the loss function, and $C$ is the constraint set in a constrained minimization

5: problem, then the proximal distance principle mandates minimizing the penalized loss $f(\bx)+\frac{\rho}{2}\dist(\bx,C)^2$ and following the solution $\bx_{\rho}$ 

6: to its limit as $\rho$ tends to $\infty$. At each iteration

7: the squared Euclidean distance $\dist(\bx,C)^2$ is majorized by the spherical quadratic

8: $\|\bx-P_C(\bx_k)\|^2$, where $P_C(\bx_k)$ denotes the projection of

9: the current iterate $\bx_k$ onto $C$.  The minimum of the surrogate

10: function $f(\bx)+\frac{\rho}{2}\|\bx-P_C(\bx_k)\|^2$

11: is given by the proximal map $\prox_{\rho^{-1}f}[P_C(\bx_k)]$.

12: The next iterate $\bx_{k+1}$ automatically decreases the original

13: penalized loss for fixed $\rho$. 

14: Since many explicit projections and proximal maps are known,

15: it is straightforward to derive and implement novel 

16: optimization algorithms in this setting. These algorithms can take hundreds if not

17: thousands of iterations to converge, but the stereotyped nature of each

18: iteration makes proximal distance algorithms competitive with traditional algorithms. For

19: convex problems, we prove global convergence. Our numerical examples

20: include a) linear programming, b) nonnegative quadratic programming,

21: c) projection to the closest kinship matrix, d) projection onto a second-order cone constraint, e) calculation of Horn's copositive matrix index, f) linear complementarity programming, and g) sparse principal components analysis. The proximal distance

22: algorithm in each case is competitive or superior in speed to 

23: traditional methods.\\

24: \svskip \\

25: \noindent {\em Key words and phrases:}  constrained optimization, EM algorithm,

26: majorization, projection, proximal map \\

27: \textit{Math Subject Classifications:} 90C59, 90C26, 65K05

28: \end{abstract}