low rank sums of squares representations
play

Low-rank sums-of-squares representations Cynthia Vinzant, North - PowerPoint PPT Presentation

Low-rank sums-of-squares representations Cynthia Vinzant, North Carolina State University joint work with Greg Blekherman, Daniel Plaumann, and Rainer Sinn JMM 2017 Cynthia Vinzant Low-rank sums-of-squares representations Sums of squares and


  1. Low-rank sums-of-squares representations Cynthia Vinzant, North Carolina State University joint work with Greg Blekherman, Daniel Plaumann, and Rainer Sinn JMM 2017 Cynthia Vinzant Low-rank sums-of-squares representations

  2. Sums of squares and nonnegative polynomials A representation of a element f ∈ R as a sum of squares over a ring R (usually R [ x 0 , . . . , x n ] or a quotient) is an expression f = h 2 1 + . . . + h 2 where h j ∈ R . r Cynthia Vinzant Low-rank sums-of-squares representations

  3. Sums of squares and nonnegative polynomials A representation of a element f ∈ R as a sum of squares over a ring R (usually R [ x 0 , . . . , x n ] or a quotient) is an expression f = h 2 1 + . . . + h 2 where h j ∈ R . r Over R = R [ x 0 , . . . , x n ], this certifies the nonnegativity of f on R n +1 . e.g. x 4 − 4 x 3 y + 5 x 2 y 2 − 2 xy 3 + y 4 Cynthia Vinzant Low-rank sums-of-squares representations

  4. Sums of squares and nonnegative polynomials A representation of a element f ∈ R as a sum of squares over a ring R (usually R [ x 0 , . . . , x n ] or a quotient) is an expression f = h 2 1 + . . . + h 2 where h j ∈ R . r Over R = R [ x 0 , . . . , x n ], this certifies the nonnegativity of f on R n +1 . e.g. x 4 − 4 x 3 y + 5 x 2 y 2 − 2 xy 3 + y 4 = ( x 2 − 2 xy ) 2 + ( xy − y 2 ) 2 Cynthia Vinzant Low-rank sums-of-squares representations

  5. Sums of squares and nonnegative polynomials A representation of a element f ∈ R as a sum of squares over a ring R (usually R [ x 0 , . . . , x n ] or a quotient) is an expression f = h 2 1 + . . . + h 2 where h j ∈ R . r Over R = R [ x 0 , . . . , x n ], this certifies the nonnegativity of f on R n +1 . e.g. x 4 − 4 x 3 y + 5 x 2 y 2 − 2 xy 3 + y 4 = ( x 2 − 2 xy ) 2 + ( xy − y 2 ) 2 Let Σ n , 2 d denote the sums of squares in R [ x 0 , . . . , x n ] 2 d and P n , 2 d denote polynomials in R [ x 0 , . . . , x n ] 2 d nonnegative on R n +1 . Cynthia Vinzant Low-rank sums-of-squares representations

  6. Sums of squares and nonnegative polynomials A representation of a element f ∈ R as a sum of squares over a ring R (usually R [ x 0 , . . . , x n ] or a quotient) is an expression f = h 2 1 + . . . + h 2 where h j ∈ R . r Over R = R [ x 0 , . . . , x n ], this certifies the nonnegativity of f on R n +1 . e.g. x 4 − 4 x 3 y + 5 x 2 y 2 − 2 xy 3 + y 4 = ( x 2 − 2 xy ) 2 + ( xy − y 2 ) 2 Let Σ n , 2 d denote the sums of squares in R [ x 0 , . . . , x n ] 2 d and P n , 2 d denote polynomials in R [ x 0 , . . . , x n ] 2 d nonnegative on R n +1 . Theorem (Hilbert): Σ n , 2 d = P n , 2 d if and only if n = 1 or 2 d = 2 or ( n , 2 d ) = (2 , 4) . Motzkin non-example: x 2 y 4 + x 4 y 2 − 3 x 2 y 2 z 2 + z 6 ∈ P 2 , 6 \ Σ 2 , 6 Cynthia Vinzant Low-rank sums-of-squares representations

  7. Number of squares n = 1 : A nonnegative bivariate form is a sum of two squares Proof: Factor f = ( p + ✐ q )( p − ✐ q ) = p 2 + q 2 where p , q ∈ R [ x 0 , x 1 ] d 2 d = 2 : A nonnegative quadratic form in P n , 2 is a sum of n + 1 squares Proof: Diagonalization of quadratic forms ( n , 2 d ) = (2 , 4) : A nonnegative ternary quartic is a sum of three squares Proof by Hilbert, 1888 Cynthia Vinzant Low-rank sums-of-squares representations

  8. Number of squares n = 1 : A nonnegative bivariate form is a sum of two squares Proof: Factor f = ( p + ✐ q )( p − ✐ q ) = p 2 + q 2 where p , q ∈ R [ x 0 , x 1 ] d 2 d = 2 : A nonnegative quadratic form in P n , 2 is a sum of n + 1 squares Proof: Diagonalization of quadratic forms ( n , 2 d ) = (2 , 4) : A nonnegative ternary quartic is a sum of three squares Proof by Hilbert, 1888 Our goal: Unify/generalize these results using varieties of minimal degree Cynthia Vinzant Low-rank sums-of-squares representations

  9. Quadratic forms on varieties Let . . . Zar , X ⊂ P N ( C ) = a real, nondegenerate irreducible variety equal to X ( R ) R [ X ] k = R [ x 0 , . . . , x N ] k / I ( X ) = coordinate ring of X in degree k , Cynthia Vinzant Low-rank sums-of-squares representations

  10. Quadratic forms on varieties Let . . . Zar , X ⊂ P N ( C ) = a real, nondegenerate irreducible variety equal to X ( R ) R [ X ] k = R [ x 0 , . . . , x N ] k / I ( X ) = coordinate ring of X in degree k , Σ X = { ℓ 2 1 + . . . + ℓ 2 r : ℓ i ∈ R [ X ] 1 } ⊂ R [ X ] 2 , and P X = { q ∈ R [ X ] 2 : q ( x ) ≥ 0 for all x ∈ X ( R ) } ⊂ R [ X ] 2 Cynthia Vinzant Low-rank sums-of-squares representations

  11. Quadratic forms on varieties Let . . . Zar , X ⊂ P N ( C ) = a real, nondegenerate irreducible variety equal to X ( R ) R [ X ] k = R [ x 0 , . . . , x N ] k / I ( X ) = coordinate ring of X in degree k , Σ X = { ℓ 2 1 + . . . + ℓ 2 r : ℓ i ∈ R [ X ] 1 } ⊂ R [ X ] 2 , and P X = { q ∈ R [ X ] 2 : q ( x ) ≥ 0 for all x ∈ X ( R ) } ⊂ R [ X ] 2 If X = ν d ( P n ), then Σ X ∼ = Σ n , 2 d and P X ∼ = P n , 2 d . Cynthia Vinzant Low-rank sums-of-squares representations

  12. Quadratic forms on varieties Let . . . Zar , X ⊂ P N ( C ) = a real, nondegenerate irreducible variety equal to X ( R ) R [ X ] k = R [ x 0 , . . . , x N ] k / I ( X ) = coordinate ring of X in degree k , Σ X = { ℓ 2 1 + . . . + ℓ 2 r : ℓ i ∈ R [ X ] 1 } ⊂ R [ X ] 2 , and P X = { q ∈ R [ X ] 2 : q ( x ) ≥ 0 for all x ∈ X ( R ) } ⊂ R [ X ] 2 If X = ν d ( P n ), then Σ X ∼ = Σ n , 2 d and P X ∼ = P n , 2 d . Ex: Let X = ν 2 ( P 1 ), where ν 2 ([ x : y ]) = [ x 2 : xy : y 2 ] = [ a : b : c ]. Then ( x 2 − 2 xy ) 2 + ( xy − y 2 ) 2 ∈ Σ 1 , 4 ↔ ( a − 2 b ) 2 + ( b − c ) 2 ∈ Σ X Cynthia Vinzant Low-rank sums-of-squares representations

  13. Quadratic forms on varieties Let . . . Zar , X ⊂ P N ( C ) = a real, nondegenerate irreducible variety equal to X ( R ) R [ X ] k = R [ x 0 , . . . , x N ] k / I ( X ) = coordinate ring of X in degree k , Σ X = { ℓ 2 1 + . . . + ℓ 2 r : ℓ i ∈ R [ X ] 1 } ⊂ R [ X ] 2 , and P X = { q ∈ R [ X ] 2 : q ( x ) ≥ 0 for all x ∈ X ( R ) } ⊂ R [ X ] 2 If X = ν d ( P n ), then Σ X ∼ = Σ n , 2 d and P X ∼ = P n , 2 d . Ex: Let X = ν 2 ( P 1 ), where ν 2 ([ x : y ]) = [ x 2 : xy : y 2 ] = [ a : b : c ]. Then ( x 2 − 2 xy ) 2 + ( xy − y 2 ) 2 ∈ Σ 1 , 4 ↔ ( a − 2 b ) 2 + ( b − c ) 2 ∈ Σ X Theorem (Blekherman-Smith-Velasco): Σ X = P X if and only if X is a variety of minimal degree (i.e. deg( X ) = codim( X ) + 1). Cynthia Vinzant Low-rank sums-of-squares representations

  14. Quadratic forms on varieties Let . . . Zar , X ⊂ P N ( C ) = a real, nondegenerate irreducible variety equal to X ( R ) R [ X ] k = R [ x 0 , . . . , x N ] k / I ( X ) = coordinate ring of X in degree k , Σ X = { ℓ 2 1 + . . . + ℓ 2 r : ℓ i ∈ R [ X ] 1 } ⊂ R [ X ] 2 , and P X = { q ∈ R [ X ] 2 : q ( x ) ≥ 0 for all x ∈ X ( R ) } ⊂ R [ X ] 2 If X = ν d ( P n ), then Σ X ∼ = Σ n , 2 d and P X ∼ = P n , 2 d . Ex: Let X = ν 2 ( P 1 ), where ν 2 ([ x : y ]) = [ x 2 : xy : y 2 ] = [ a : b : c ]. Then ( x 2 − 2 xy ) 2 + ( xy − y 2 ) 2 ∈ Σ 1 , 4 ↔ ( a − 2 b ) 2 + ( b − c ) 2 ∈ Σ X Theorem (Blekherman-Smith-Velasco): Σ X = P X if and only if X is a variety of minimal degree (i.e. deg( X ) = codim( X ) + 1). ν d ( P n ) has minimal degree ⇔ n = 1, d = 1, or ( n , d ) = (2 , 2) Corollary: Hilbert’s result. Cynthia Vinzant Low-rank sums-of-squares representations

  15. Varieties of minimal degree Theorem: If X is a variety of minimal degree, then any q ∈ P X is a sum of dim( X ) + 1 squares. For generic q this bound is tight. Cynthia Vinzant Low-rank sums-of-squares representations

  16. Varieties of minimal degree Theorem: If X is a variety of minimal degree, then any q ∈ P X is a sum of dim( X ) + 1 squares. For generic q this bound is tight. A variety of minimal degree is isomorphic to one of the following: ◮ a quadratic hypersurface ◮ ν d ( P 1 ) ◮ ν 2 ( P 2 ) ◮ a rational normal scroll ◮ a cone over one of the above. Cynthia Vinzant Low-rank sums-of-squares representations

  17. Varieties of minimal degree Theorem: If X is a variety of minimal degree, then any q ∈ P X is a sum of dim( X ) + 1 squares. For generic q this bound is tight. A variety of minimal degree is isomorphic to one of the following: ◮ a quadratic hypersurface ◮ ν d ( P 1 ) ⇒ f ∈ P 1 , 2 d = sum of 2 squares ◮ ν 2 ( P 2 ) ◮ a rational normal scroll ◮ a cone over one of the above. Cynthia Vinzant Low-rank sums-of-squares representations

  18. Varieties of minimal degree Theorem: If X is a variety of minimal degree, then any q ∈ P X is a sum of dim( X ) + 1 squares. For generic q this bound is tight. A variety of minimal degree is isomorphic to one of the following: ◮ a quadratic hypersurface ◮ ν d ( P 1 ) ⇒ f ∈ P 1 , 2 d = sum of 2 squares ◮ ν 2 ( P 2 ) ⇒ f ∈ P 2 , 4 = sum of 3 squares ◮ a rational normal scroll ◮ a cone over one of the above. Cynthia Vinzant Low-rank sums-of-squares representations

  19. Varieties of minimal degree Theorem: If X is a variety of minimal degree, then any q ∈ P X is a sum of dim( X ) + 1 squares. For generic q this bound is tight. A variety of minimal degree is isomorphic to one of the following: ◮ a quadratic hypersurface ◮ ν d ( P 1 ) ⇒ f ∈ P 1 , 2 d = sum of 2 squares ◮ ν 2 ( P 2 ) ⇒ f ∈ P 2 , 4 = sum of 3 squares ◮ a rational normal scroll ⇒ ?? ◮ a cone over one of the above. Cynthia Vinzant Low-rank sums-of-squares representations

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend