The permanent of a matrix is a quantity similar to the determinant but without any minus signs. It shows up in combinatorics and theoretical computer science, and computing it exactly is notoriously hard. The two best-known algorithms for this, due to Ryser and Glynn, both work by evaluating a certain product over a large collection of points and then combining the results with appropriate weights. The paper shows that both algorithms are actually special cases of a single unifying principle: pick a set of evaluation points and a weighting scheme, then check a simple orthogonality condition to confirm the weighted sum recovers the permanent.
Using this framework, the authors introduce a third algorithm based on the discrete Fourier transform, where the permanent appears as one coefficient of a polynomial computed over a cyclic group of numbers. This new scheme runs in exactly the same time as the classical algorithms and can be made exact over finite fields using a standard number theory trick called the Chinese remainder theorem. The paper also explains why Ryser's and Glynn's approaches satisfy the orthogonality condition almost automatically, while the Fourier-based approach requires a careful arithmetic condition on which evaluation points are used. All three methods turn out to be examples of character sums over finite abelian groups, giving a clean algebraic picture that unifies previously separate results.