l
play

L d d rank d L d - PowerPoint PPT Presentation

Jul 22, 2023 •177 likes •681 views

Ranks of matrices with few distinct entries Boris Bukh 8 August 2017 d L d d rank d L d d A special case: equiangular lines Family L of lines in R d is

  1. Ranks of matrices with few distinct entries Boris Bukh 8 August 2017  d  ∈ L d     d   rank   d     ∈ L d   d

  2. A special case: equiangular lines Family L of lines in R d is equiangular when all pairwise angles ∡ ℓℓ ′ are equal, for ℓ, ℓ ′ ∈ L Examples: d = 2 d = 3 (Large diagonals)

  3. Gram matrices Lines l 1 , . . . , L n in R d = ⇒ Unit vectors v 1 , . . . , v n in R d (line directions) = ⇒ � � Matrix of inner products � v i , v j � i , j (Gram matrix)

  4. Gram matrices Lines l 1 , . . . , L n in R d Equiangular = = ⇒ ⇒ Unit vectors v 1 , . . . , v n in R d ???? (line directions) = ⇒ � � Matrix of inner products � v i , v j � i , j (Gram matrix)

  5. Gram matrices Lines l 1 , . . . , L n in R d Equiangular = = ⇒ ⇒ Unit vectors v 1 , . . . , v n in R d � v i , v j � ∈ {− α, + α } (line directions) = = ⇒ ⇒ ± α   1 � � Matrix of inner products � v i , v j � 1   i , j   1 (Gram matrix)   ± α   1

  6. Gram matrices Lines l 1 , . . . , L n in R d Equiangular ⇐ ⇐ ⇒ ⇒ Unit vectors v 1 , . . . , v n in R d � v i , v j � ∈ {− α, + α } (line directions) ⇐ ⇐ ⇒ ⇒ ± α   1 � � Matrix of inner products � v i , v j � 1   i , j   1 (Gram matrix)   ± α   Positive semidefinite 1

  7. Gram matrices Unit vectors v 1 , v 2 , . . . , v n in R d n vectors = ⇒   A =  v 1 v 2 · · · v n Rank ≤ d    = ⇒ Gram matrix M = A T A Rank ≤ d

  8. General problem How small can a rank of an ( L , d )-matrix be? General ( L , d )-matrix  d  ∈ L d     d     d     ∈ L d   d

  9. General problem How small can a rank of an ( L , d )-matrix be? General ( L , d )-matrix  d  ∈ L d     d     d     ∈ L d   d If M is an ( L , d )-matrix, then M − dJ is ( L − d , 0)-matrix of almost the same rank. So, with little loss we may assume that d = 0.

  10. Special case: graph eigenvalues What is the maximum eigenvalue multiplicity of λ ? Details: Number λ is fixed We consider adjacency matrices of graphs on n vertices We seek the graph that maximizes the multiplicity of eigenvalue λ

  11. Special case: graph eigenvalues What is the maximum eigenvalue multiplicity of λ ? General adjacency matrix: :  0  { 0 , 1 } 0     0     0    { 0 , 1 }  0   0

  12. Special case: graph eigenvalues What is the maximum eigenvalue multiplicity of λ ? Multiplicity of λ in a Nullity of a ⇐ ⇒ general adjacency matrix: matrix of the form:  0   − λ  { 0 , 1 } { 0 , 1 } 0 − λ         0 − λ         0 − λ      { 0 , 1 }   { 0 , 1 }  0 − λ     0 − λ Rank + nullity = n

  13. ( L , d ) -matrices: some examples Equiangular lines Multiplicity of graph eigenvalues Sets in R d with few distances Set systems with restricted intersection

  14. ( L , d ) -matrices: some examples Equiangular lines Multiplicity of graph eigenvalues Sets in R d with few distances Set systems with restricted intersection S 1 , . . . , S n are d -element sets with | S i ∩ S j | ∈ L v 1 , . . . , v n are characteristic vectors   A =  v 1 v 2 · · · v n  is made of 0’s and 1’s   M = A T A is an ( L , d )-matrix

  15. L -matrices: the upper bound General L -matrix  0  ∈ L 0     0     0     ∈ L 0   0 “Polynomial method” (Koornwinder? Frankl–Wilson?) Suppose | L | = k and 0 �∈ L , and M is an n -by- n L -matrix of rank r . Then � r + k � n ≤ . k

  16. L -matrices: the upper bound General L -matrix  0  ∈ L 0     0     0     ∈ L 0   0 “Polynomial method” (Koornwinder? Frankl–Wilson?) Suppose | L | = k and 0 �∈ L , and M is an n -by- n Sharp for some sets L L -matrix of rank r . Then � r + k � n ≤ . k

  17. An example  0  { 1 , 3 } 0     0     0    { 1 , 3 }  0   0 ⇒ size at most O ( r 2 ) Polynomial method: rank r =

  18. An example  0  { 1 , 3 } 0     0     0    { 1 , 3 }  0   0 ⇒ size at most O ( r 2 ) Polynomial method: rank r = Modulo 2: almost full rank, size at most r + 1

  19. General results N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . Theorem (B.) For a set L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > r for some r 2 There is an integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 3 lim r →∞ N ( r , L ) / r exists and is > 1

  20. General results N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . Theorem (B.) For a set L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > kr for some r linear 2 There is a integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 3 N ( r , L ) = Ω( r 3 / 2 )

  21. Corollaries for the special case G ( n , λ ) = max { mult. λ in a n -vertex graph } D ( n , λ ) = max { mult. λ in a n -vertex digraph } Theorem (B.) 1 If λ is an algebraic integer of degree d, then D ( n , λ ) = n / d − O ( √ n ) . 2 Otherwise, λ is not an eigenvalue of any { 0 , 1 } -matrix Graph eigenvalues: Same holds for G ( n , λ ) if degree of λ is at most 4 The general case is open

  22. Mathematics is beautiful!

  23. Proofs: algebraic reason N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . For L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > r for some r 2 There is an integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 r →∞ N ( r , L ) / r exists and is > 1 lim 3

  24. Proofs: algebraic reason N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . For L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > r for some r 2 There is an integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 Proof of 1 = ⇒ 2 . Assume M is an L -matrix of size n . def Let P n ( α 1 , . . . , α n ) = det M , homogeneous of degree n .  0 α 1 ··· α 3  α 2 0 ··· α 1 P n ( α 1 , . . . , α k ) = det . . . ... . . . .   . . α 1 α 1 ··· 0

  25. Proofs: algebraic reason N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . For L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > r for some r 2 There is an integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 Proof of 1 = ⇒ 2 . Assume M is an L -matrix of size n . def Let P n ( α 1 , . . . , α n ) = det M , homogeneous of degree n . � 0 1 ··· 1 � 1 0 ··· 1 = ( − 1) n − 1 ( n − 1) P n (1 , . . . , 1) = det . . . . ... . . . . . 1 1 ··· 0

  26. Proofs: algebraic reason N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . For L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > r for some r 2 There is an integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 Proof of 1 = ⇒ 2 . Assume M is an L -matrix of size n , M ′ is a submatrix of size n − 1 def Let P n ( α 1 , . . . , α n ) = det M , homogeneous of degree n . def Let P n − 1 ( α 1 , . . . , α n ) = det M ′ , homogeneous of degree n − 1. P n (1 , . . . , 1) = ( − 1) n − 1 ( n − 1) P n − 1 (1 , . . . , 1) = ( − 1) n − 2 ( n − 2)

  27. Proofs: algebraic reason N ( r , L ) = max { n : there is an n -by- n L -matrix of rank ≤ r } . For L = { α 1 , . . . , α k } , the following are equivalent 1 N ( r − 1 , L ) > r for some r 2 There is an integer homogeneous polynomial P s.t. P ( α 1 , . . . , α k ) = 0 and P (1 , 1 , . . . , 1) = 1 Proof of 1 = ⇒ 2 . Assume M is an L -matrix of size n , M ′ is a submatrix of size n − 1 def Let P n ( α 1 , . . . , α n ) = det M , homogeneous of degree n . def Let P n − 1 ( α 1 , . . . , α n ) = det M ′ , homogeneous of degree n − 1. P n (1 , . . . , 1) = ( − 1) n − 1 ( n − 1) � ⇒ P = ( P n − α 1 P n − 1 ) 2 = P n − 1 (1 , . . . , 1) = ( − 1) n − 2 ( n − 2) P ( α 1 , . . . , α k ) = 0

Recommend

More recommend