Positive semidefinite rank

Transcription

1 1/15 Positive semidefinite rank Hamza Fawzi (MIT, LIDS) Joint work with João Gouveia (Coimbra), Pablo Parrilo (MIT), Richard Robinson (Microsoft), James Saunderson (Monash), Rekha Thomas (UW) DIMACS Workshop on Distance Geometry July 2016

2 2/15 Euclidean distance matrices Theorem (Schoenberg, 1935) M is an Euclidean distance matrix if and only if diag(m) = 0 and [M 1,i + M 1,j M i,j ] 2 i,j n is positive semidefinite. Allows us to express certain distance geometry problems as semidefinite programs

3 2/15 Euclidean distance matrices Theorem (Schoenberg, 1935) M is an Euclidean distance matrix if and only if diag(m) = 0 and [M 1,i + M 1,j M i,j ] 2 i,j n is positive semidefinite. Allows us to express certain distance geometry problems as semidefinite programs Which convex sets can be represented using semidefinite programming?

4 3/15 Semidefinite representation Feasible set of a semidefinite program: { X 0 (positive semidefinite constraint) A(X ) = b (linear constraints)

5 3/15 Semidefinite representation Feasible set of a semidefinite program: { X 0 (positive semidefinite constraint) A(X ) = b (linear constraints) Convex set C has a semidefinite representation of size d if: C = π(s d + L) S d + = d d positive semidefinite matrices L = affine subspace π = linear map L S d + π C

6 4/15 Examples of semidefinite representations Examples: EDM n+1 has SDP representation of size n

7 4/15 Examples of semidefinite representations Examples: EDM n+1 has SDP representation of size n Disk in R 2 has a SDP representation of size 2 [ ] x 2 + y 2 1 x y 1 0 y 1 + x

8 4/15 Examples of semidefinite representations Examples: EDM n+1 has SDP representation of size n Disk in R 2 has a SDP representation of size 2 [ ] x 2 + y 2 1 x y 1 0 y 1 + x Square [ 1, 1] 2 has a SDP representation of size 3 [ 1, 1] 2 = (x1, x2) R2 : u R 1 x 1 x 2 x 1 1 u 0 x 2 u 1

9 5/15 Existential question vs. complexity question Existential question: Which convex sets admit a semidefinite representation? Helton-Nie conjecture: Any convex set defined using polynomial inequalities has a semidefinite representation

10 5/15 Existential question vs. complexity question Existential question: Which convex sets admit a semidefinite representation? Helton-Nie conjecture: Any convex set defined using polynomial inequalities has a semidefinite representation Complexity question: Given a convex set C, what is smallest semidefinite representation of C?

11 5/15 Existential question vs. complexity question Existential question: Which convex sets admit a semidefinite representation? Helton-Nie conjecture: Any convex set defined using polynomial inequalities has a semidefinite representation Complexity question: Given a convex set C, what is smallest semidefinite representation of C? Positive semidefinite rank

12 Importance of lifting 6/15

13 Importance of lifting 6/15

14 Importance of lifting 6/15

15 6/15 Importance of lifting Ben-Tal and Nemirovski: Regular polygon with 2 n sides can be described using only n inequalities!

16 6/15 Importance of lifting Ben-Tal and Nemirovski: Regular polygon with 2 n sides can be described using only n inequalities! Lift = inverse of elimination (cf. Pablo s talk)

17 7/15 Lifts of polytopes and ranks of matrices P polytope in R d Slack matrix of P: Nonnegative matrix M of size #facets(p) #vertices(p): where M i,j = h i gi T v j gi T x h i are the facet inequalities of P v j are the vertices of P v j h i g T i v j g T i x h i

18 7/15 Lifts of polytopes and ranks of matrices P polytope in R d Slack matrix of P: Nonnegative matrix M of size #facets(p) #vertices(p): where M i,j = h i gi T v j gi T x h i are the facet inequalities of P v j are the vertices of P v j h i g T i v j g T i x h i

19 7/15 Lifts of polytopes and ranks of matrices P polytope in R d Slack matrix of P: Nonnegative matrix M of size #facets(p) #vertices(p): where M i,j = h i gi T v j gi T x h i are the facet inequalities of P v j are the vertices of P v j h i g T i v j g T i x h i x

20 7/15 Lifts of polytopes and ranks of matrices P polytope in R d Slack matrix of P: Nonnegative matrix M of size #facets(p) #vertices(p): where M i,j = h i gi T v j gi T x h i are the facet inequalities of P v j are the vertices of P v j h i g T i v j g T i x h i

21 7/15 Lifts of polytopes and ranks of matrices P polytope in R d Slack matrix of P: Nonnegative matrix M of size #facets(p) #vertices(p): where M i,j = h i gi T v j gi T x h i are the facet inequalities of P v j are the vertices of P v j h i g T i v j g T i x h i

22 7/15 Lifts of polytopes and ranks of matrices P polytope in R d Slack matrix of P: Nonnegative matrix M of size #facets(p) #vertices(p): where M i,j = h i gi T v j gi T x h i are the facet inequalities of P v j are the vertices of P v j h i g T i v j g T i x h i

23 8/15 Positive semidefinite rank M R p q with nonnegative entries Positive semidefinite factorization: M ij = Tr(A i B j ), where A i, B j S k + rank psd (M) = size of smallest psd factorization B j A i Tr(A i B j )

24 9/15 Example Consider M ij = (i j) 2 for 1 i, j n: M = rank psd (M) = 2 (independent of n): Let A i = [ ] [ ] [ ] T 1 i 1 1 i i 2 = and B i i j = One can verify that M ij = Tr(A j B j ). [ ] j 2 j = j 1 [ j 1 ] [ ] T j. 1

25 10/15 SDP representations and psd rank Theorem (Gouveia, Parrilo, Thomas, 2011) Let P be polytope with slack matrix M. The smallest semidefinite representation of P has size exactly rank psd (M). Polytope P Slack matrix M rank psd (M) Works for more generally for convex sets (slack matrix is infinite) Proof based on duality for semidefinite programming

26 10/15 SDP representations and psd rank Theorem (Gouveia, Parrilo, Thomas, 2011) Let P be polytope with slack matrix M. The smallest semidefinite representation of P has size exactly rank psd (M). Polytope P Slack matrix M rank psd (M) Works for more generally for convex sets (slack matrix is infinite) Proof based on duality for semidefinite programming Example: Slack matrix of square [ 1, 1] 2 has positive semidefinite rank 3.

27 11/15 Properties of rank psd Satisfies the usual properties one would expect for a rank (invariance under scaling, subadditivity, etc.) [Fawzi, Gouveia, Parrilo, Robinson, Thomas, Positive semidefinite rank, Math. Prog., 2015] Connection with problems in information theory NP-hard to compute [Shitov, 2016]

28 12/15 Linear programming (LP) lifts Polytope P has LP lift of size d if it can be written as P = π(r d + L) where L affine subspace and π linear map Nonnegative factorization of M of size d: M ij = a T i b j where a i, b j R d + rank + (M) := size of smallest nonnegative factorization of M Theorem (Yannakakis, 1991) Let P be polytope with slack matrix M. The smallest LP lift of P has size exactly rank + (M).

29 13/15 LP lifts vs. SDP lifts Example The square P = [ 1, 1] 2 : SDP lifts: P has an SDP lift of size 3: [ 1, 1] 2 = (x1, x2) R2 : u R 1 x 1 x 2 x 1 1 u 0 x 2 u 1 SDP lift of size 3. LP lifts: Can show that any LP lift of [ 1, 1] 2 must have size 4. Stable set polytope for perfect graphs: SDP lift of linear size (Lovász) but no currently known LP lift of polynomial size

30 14/15 LP lifts vs. SDP lifts Question: How powerful are SDP lifts compared to LP lifts? Theorem (Fawzi, Saunderson, Parrilo, 2015) There is a family of polytopes P d R 2d such that ( ) rank psd (P d ) log d rank + (P d ) O 0. d P d = trigonometric cyclic polytope (generalization of regular polygons) Construction uses tools from Fourier analysis + sparse sums of squares

31 15/15 Conclusion Semidefinite representations of convex sets Connection with matrix factorization Linear programming vs. semidefinite programming lifts for polytopes For more information: [Fawzi, Gouveia, Parrilo, Robinson, Thomas, Positive semidefinite rank, Math. Prog., 2015]

32 15/15 Conclusion Semidefinite representations of convex sets Connection with matrix factorization Linear programming vs. semidefinite programming lifts for polytopes For more information: [Fawzi, Gouveia, Parrilo, Robinson, Thomas, Positive semidefinite rank, Math. Prog., 2015] Thank you!