TOPICS
Search

Glynn Formula


The Glynn formula expresses the permanent of a matrix as a sum over sign vectors. For an n×n matrix A=(a_(ij)) over a field whose characteristic is not 2, with n a positive integer, it states that

 per(A)=1/(2^(n-1))sum_(delta)(product_(i=1)^ndelta_i)product_(j=1)^n(sum_(i=1)^ndelta_ia_(ij)),

where the sum is over the 2^(n-1) vectors delta=(delta_1,...,delta_n) in {-1,1}^n with delta_1=1 (Glynn 2010).

To prove the formula, expand the product of column sums and let lambda_i count the factors drawn from row i. Summing over signs cancels every term unless each lambda_i with i>1 is odd. Since sum_(i)lambda_i=n, this forces every lambda_i to equal 1. The surviving terms are exactly the permutation terms defining the permanent, each with weight 2^(n-1).

Ordering the sign choices by a Gray code changes a single sign at each step, so the n column sums can be updated in O(n) arithmetic operations. Forming their product also takes O(n) operations per term, giving O(n2^n) arithmetic complexity, the same asymptotic bound as the implementation of the Ryser formula using a Gray code. Here O denotes big-O notation.


See also

Gray Code, Permanent, Ryser Formula

Explore with Wolfram|Alpha

References

Glynn, D. G. "The Permanent of a Square Matrix." Europ. J. Combin. 31, 1887-1891, 2010. https://doi.org/10.1016/j.ejc.2010.01.010.

Cite this as:

Weisstein, Eric W. "Glynn Formula." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/GlynnFormula.html

Subject classifications