1: \begin{abstract} 

2:   We investigate the representation of symmetric polynomials as a sum

3:   of squares. Since this task is solved using semidefinite programming

4:   tools we explore the geometric, algebraic, and computational

5:   implications of the presence of discrete symmetries in semidefinite

6:   programs.  It is shown that symmetry exploitation allows a

7:   significant reduction in both matrix size and number of decision

8:   variables.  This result is applied to semidefinite programs arising

9:   from the computation of sum of squares decompositions for

10:   multivariate polynomials. The results, reinterpreted from an

11:   invariant-theoretic viewpoint, provide a novel representation of a

12:   class of nonnegative symmetric polynomials.  The main theorem states

13:   that an invariant sum of squares polynomial is a sum of inner

14:   products of pairs of matrices, whose entries are invariant

15:   polynomials. In these pairs, one of the matrices is computed based

16:   on the real irreducible representations of the group, and the other

17:   is a sum of squares matrix. The reduction techniques enable the

18:   numerical solution of large-scale instances, otherwise

19:   computationally infeasible to solve.

20: 

21: \noindent 

22: {\bf AMS Subject Classification:} 14P99, 26C05, 13A50, 68W30, 90C22 

23: %

24: %

25: %

26: %

27: %

28: %

29: %

30: \end{abstract}